Merge remote-tracking branch 'origin/main' into HEAD

; Conflicts:
;	ROADMAP.md

.github/workflows/release.yml

@@ -18,7 +18,14 @@ concurrency:
   cancel-in-progress: false
 
 jobs:
+  # Regenerates contributors/contributors.json for the release commit; every build bundles it.
+  credits:
+    permissions:
+      contents: read
+      pull-requests: read
+    uses: storytold/craft-common-actions/.github/workflows/refresh-contributors.yml@main
   release:
+    needs: credits
     # Only this job publishes the GitHub release, so only it gets write access.
     permissions:
       contents: write
@@ -31,6 +38,7 @@ jobs:
       craft-fonts-ref: b076417b5880bd4154bd22f00695b1d28c6bea83
       web-craft-fonts: true
       version: ${{ inputs.version }}
+      contributors-artifact: ${{ needs.credits.outputs.artifact-name }}
     secrets:
       APPLE_CERTIFICATE: ${{ secrets.APPLE_CERTIFICATE }}
       APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}

Cargo.lock

@@ -1158,10 +1158,21 @@ dependencies = [
  "zip 9.1.0",
 ]
 
+[[package]]
+name = "craft-web-api"
+version = "0.1.0"
+source = "git+https://github.com/storytold/craft-libs?rev=fd5e2feff6ea25bed371a99cc732b3968d621505#fd5e2feff6ea25bed371a99cc732b3968d621505"
+dependencies = [
+ "js-sys",
+ "serde_json",
+ "wasm-bindgen",
+ "wasm-bindgen-futures",
+]
+
 [[package]]
 name = "craft-web-host"
 version = "0.1.0"
-source = "git+https://github.com/storytold/craft-libs?rev=2e2312ddaf21bc3d85cd3395d0955c7d239306f6#2e2312ddaf21bc3d85cd3395d0955c7d239306f6"
+source = "git+https://github.com/storytold/craft-libs?rev=fd5e2feff6ea25bed371a99cc732b3968d621505#fd5e2feff6ea25bed371a99cc732b3968d621505"
 dependencies = [
  "wasm-bindgen",
  "web-sys",
@@ -6795,7 +6806,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "arboard",
  "chrono",
@@ -6833,7 +6844,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-cli"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "serde_json",
  "sys-locale",
@@ -6846,11 +6857,11 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-control-key"
-version = "0.16.0"
+version = "0.15.1"
 
 [[package]]
 name = "wordcraft-doc"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "base64 0.23.1",
  "log",
@@ -6864,7 +6875,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-docbin"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "cfb",
  "chrono",
@@ -6881,7 +6892,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-docx"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "aes",
  "base64 0.23.1",
@@ -6906,7 +6917,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-engine"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "cfb",
  "chrono",
@@ -6939,7 +6950,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-fonts"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "brotli-decompressor",
  "harfrust 0.13.3",
@@ -6953,7 +6964,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-formats"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "aes",
  "base64 0.23.1",
@@ -6980,7 +6991,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-geom"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "image",
  "serde",
@@ -6988,7 +6999,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-layout"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "craft-proof",
  "image",
@@ -7008,7 +7019,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-mcp"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "log",
  "serde_json",
@@ -7020,7 +7031,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-metafile"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "kurbo",
  "thiserror 2.0.21",
@@ -7028,7 +7039,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-pdf"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "chrono",
  "craft-pdf",
@@ -7049,7 +7060,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-render"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "image",
  "log",
@@ -7063,7 +7074,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-ui-egui"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "accesskit_consumer",
  "cfb",
@@ -7096,8 +7107,9 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-web"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
+ "craft-web-api",
  "craft-web-host",
  "eframe",
  "egui",
@@ -7116,7 +7128,7 @@ dependencies = [
 
 [[package]]
 name = "wordcraft-zotero"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "log",
  "proptest",
@@ -7279,7 +7291,7 @@ checksum = "9440ea3e5aeabb0ac63af70daf835274065238cdd0cec83418f417eae38bacee"
 
 [[package]]
 name = "xtask"
-version = "0.16.0"
+version = "0.15.1"
 dependencies = [
  "serde_json",
 ]

Cargo.toml

@@ -4,7 +4,7 @@ members = ["crates/*", "apps/*", "xtask"]
 default-members = ["crates/*", "apps/wordcraft", "apps/wordcraft-cli", "xtask"]
 
 [workspace.package]
-version = "0.16.0"
+version = "0.15.1"
 edition = "2024"
 license = "MIT OR Apache-2.0"
 rust-version = "1.95"

apps/wordcraft-cli/src/main.rs

@@ -98,8 +98,8 @@ fn open(path: &Path, password: Option<&str>) -> Result<Session, String> {
     Ok(s)
 }
 
-/// The commands that open a file, and which of their arguments not starting with `--` it is (as
-/// `pos` counts). `run` opens the value of `--file`.
+/// The commands that open a file, and which positional argument it is, excluding option values.
+/// `run` opens the value of `--file`.
 const OPENS_FILE: &[(&str, usize)] = &[("convert", 0), ("info", 0), ("text", 0), ("inspect", 0), ("render", 0), ("zotero", 1)];
 
 /// The arguments as text. Only the file the command opens may have a name that isn't valid Unicode,
@@ -108,8 +108,14 @@ const OPENS_FILE: &[(&str, usize)] = &[("convert", 0), ("info", 0), ("text", 0),
 fn text_args<S: AsRef<OsStr>>(raw: &[S]) -> Result<(Vec<String>, Option<PathBuf>), String> {
     let args: Vec<String> = raw.iter().map(|a| a.as_ref().to_string_lossy().into_owned()).collect();
     let cmd = args.first().map(String::as_str).unwrap_or("");
+    let specs = match cmd {
+        "convert" => CONVERT_OPTIONS,
+        "render" => RENDER_OPTIONS,
+        "zotero" => ZOTERO_OPTIONS,
+        _ => PASSWORD_OPTIONS,
+    };
     let file_at = match OPENS_FILE.iter().find(|(c, _)| *c == cmd) {
-        Some((_, n)) => args.iter().enumerate().skip(1).filter(|(_, a)| !a.starts_with("--")).nth(*n).map(|(i, _)| i),
+        Some((_, n)) => positional_indices(args.get(1..).unwrap_or_default(), specs).get(*n).map(|i| i + 1),
         None if cmd == "run" => args.iter().position(|a| a == "--file").map(|i| i + 1),
         None => None,
     };
@@ -145,8 +151,13 @@ fn password_from(args: &[String]) -> Result<Option<String>, String> {
 }
 
 /// The positional arguments, skipping every option's value so an option value is never read as a
-/// path. Mirrors [`validate_options`], which has already checked the options are known.
+/// path. Uses the same option specifications as [`validate_options`].
 fn positionals<'a>(args: &'a [String], specs: &[OptionSpec]) -> Vec<&'a String> {
+    positional_indices(args, specs).into_iter().filter_map(|i| args.get(i)).collect()
+}
+
+/// Keep the original indices too, so [`text_args`] can open a positional path by its native name.
+fn positional_indices(args: &[String], specs: &[OptionSpec]) -> Vec<usize> {
     let mut out = Vec::new();
     let mut i = 0;
     while let Some(arg) = args.get(i) {
@@ -156,7 +167,7 @@ fn positionals<'a>(args: &'a [String], specs: &[OptionSpec]) -> Vec<&'a String>
             i += 2;
         } else {
             if !arg.starts_with("--") {
-                out.push(arg);
+                out.push(i);
             }
             i += 1;
         }

apps/wordcraft-cli/src/tests.rs

@@ -211,6 +211,10 @@ fn only_the_opened_file_may_be_other_than_unicode() {
     let ok = |raw: Vec<OsString>| text_args(&raw).unwrap().1;
     assert_eq!(ok(vec!["zotero".into(), "refresh".into(), latin1()]), Some(PathBuf::from(latin1())));
     assert_eq!(ok(vec!["render".into(), latin1(), "out.png".into(), "--page".into(), "2".into()]), Some(PathBuf::from(latin1())));
+    assert_eq!(ok(vec!["render".into(), "--page".into(), "2".into(), latin1(), "out.png".into()]), Some(PathBuf::from(latin1())));
+    assert_eq!(ok(vec!["convert".into(), "--format".into(), "flatopc".into(), latin1(), "out.xml".into()]), Some(PathBuf::from(latin1())));
+    assert_eq!(ok(vec!["info".into(), "--password".into(), "secret".into(), latin1()]), Some(PathBuf::from(latin1())));
+    assert_eq!(ok(vec!["zotero".into(), "--port".into(), "23119".into(), "refresh".into(), latin1()]), Some(PathBuf::from(latin1())));
     assert_eq!(ok(vec!["run".into(), "--print".into(), "--file".into(), latin1()]), Some(PathBuf::from(latin1())));
     assert_eq!(ok(vec!["info".into(), "in.docx".into()]), None);
     // Every command that opens a file reaches the open with it (no such file here), so
@@ -228,6 +232,7 @@ fn only_the_opened_file_may_be_other_than_unicode() {
     for (raw, n) in [
         (vec!["zotero".into(), latin1(), "in.docx".into()], 2),
         (vec!["render".into(), "in.docx".into(), latin1()], 3),
+        (vec!["render".into(), "--page".into(), latin1(), "in.docx".into(), "out.png".into()], 3),
         (vec!["run".into(), "--save".into(), latin1()], 3),
         (vec!["commands".into(), latin1()], 2),
     ] {

apps/wordcraft-cli/tests/it/non_utf8_args.rs

@@ -53,6 +53,15 @@ fn the_file_a_command_opens_may_have_any_name() {
     let _ = std::fs::remove_dir_all(dir);
 }
 
+#[test]
+fn an_option_before_the_input_keeps_its_native_name() {
+    let (dir, file) = folder("option-first");
+    let saved = dir.join("out.xml");
+    ok(&cli(&["convert".into(), "--format".into(), "flatopc".into(), file, saved.clone().into()]));
+    assert!(std::fs::read_to_string(saved).unwrap().contains("Bonjour"));
+    let _ = std::fs::remove_dir_all(dir);
+}
+
 #[test]
 fn any_other_such_argument_is_an_error_naming_it() {
     let (dir, file) = folder("error");

apps/wordcraft-web/Cargo.toml

@@ -17,7 +17,9 @@ path = "src/main.rs"
 wordcraft-engine = { workspace = true }
 wordcraft-ui-egui = { workspace = true }
 # The host messages (storytold/craft-libs): one contract and one bridge for the web builds.
-craft-web-host = { git = "https://github.com/storytold/craft-libs", rev = "2e2312ddaf21bc3d85cd3395d0955c7d239306f6", version = "0.1.0" }
+# The page command API (storytold/craft-libs#413): window.wordcraft over the control channel.
+craft-web-api = { git = "https://github.com/storytold/craft-libs", rev = "fd5e2feff6ea25bed371a99cc732b3968d621505", version = "0.1.0" }
+craft-web-host = { git = "https://github.com/storytold/craft-libs", rev = "fd5e2feff6ea25bed371a99cc732b3968d621505", version = "0.1.0" }
 egui = { workspace = true }
 log = { workspace = true }
 serde_json = { workspace = true }

apps/wordcraft-web/src/api.rs

@@ -0,0 +1,61 @@
+//! `window.wordcraft`: the page command API (`craft-web-api`, storytold/craft-libs#413) over the
+//! app's own control channel, the one the desktop control server drives (`docs/control-protocol.md`).
+//!
+//! The crate owns the JavaScript side (`request`, `execute`, `commands`, `docInspect`, `uiInspect`,
+//! `screenshot`, as promises with a timeout). This module is WordCraft's part: a [`Request`] that
+//! queues a [`ControlRequest`] for the app, which answers it on its next frame, and waits for the
+//! reply. WordCraft's control request has no abandon state, so a call the page gave up on still runs
+//! when the app reaches it, as it does for a desktop client that timed out.
+
+use std::rc::Rc;
+use std::sync::mpsc::{Receiver, TryRecvError, channel};
+
+use craft_web_api::methods::Methods;
+use craft_web_api::page::{PageApi, Request};
+use serde_json::{Value, json};
+use wasm_bindgen_futures::JsFuture;
+use wordcraft_ui_egui::ControlRequest;
+
+/// The app id in `window.<app>`.
+const APP: &str = "wordcraft";
+
+/// How often a waiting call looks for the app's reply, in milliseconds (about a frame).
+const POLL_MS: i32 = 16;
+
+/// Put `window.wordcraft` on the page and return the queue the app drains each frame (hand it to
+/// `WordApp::with_control`). An existing `window.wordcraft` keeps the functions it already has.
+pub fn install(ctx: egui::Context) -> Receiver<ControlRequest> {
+    let (queue, requests) = channel::<ControlRequest>();
+    let request: Request = Rc::new(move |method, params| {
+        let (control, reply) = ControlRequest::new(method, params);
+        if queue.send(control).is_err() {
+            return Box::pin(async { json!({"ok": false, "error": "the app is not running"}) });
+        }
+        ctx.request_repaint();
+        Box::pin(wait(reply))
+    });
+    if let Err(e) = Rc::new(PageApi::new(APP, Methods::default(), request)).install() {
+        log::warn!("couldn't install window.{APP}: {e:?}");
+    }
+    requests
+}
+
+/// The app's reply, polled about once a frame.
+async fn wait(reply: Receiver<Value>) -> Value {
+    loop {
+        match reply.try_recv() {
+            Ok(reply) => return reply,
+            Err(TryRecvError::Empty) => sleep(POLL_MS).await,
+            Err(TryRecvError::Disconnected) => return json!({"ok": false, "error": "the app dropped the request"}),
+        }
+    }
+}
+
+async fn sleep(ms: i32) {
+    let promise = js_sys::Promise::new(&mut |resolve, _| {
+        if let Some(window) = web_sys::window() {
+            let _ = window.set_timeout_with_callback_and_timeout_and_arguments_0(&resolve, ms);
+        }
+    });
+    let _ = JsFuture::from(promise).await;
+}

apps/wordcraft-web/src/main.rs

@@ -20,6 +20,8 @@ mod prefs;
 #[cfg(any(target_arch = "wasm32", test))]
 mod save_fallback;
 
+#[cfg(target_arch = "wasm32")]
+mod api;
 #[cfg(target_arch = "wasm32")]
 mod host;
 #[cfg(target_arch = "wasm32")]

apps/wordcraft-web/src/web.rs

@@ -76,7 +76,8 @@ pub fn start() {
                     if let Some(bridge) = &app_bridge {
                         bridge.connect(&mut services, inbox.clone(), cc.egui_ctx.clone());
                     }
-                    let mut app = WordApp::new(session, services);
+                    let control = crate::api::install(cc.egui_ctx.clone());
+                    let mut app = WordApp::new(session, services).with_control(control);
                     // Pasted pictures wait apart from dropped files: the app inserts one only when the page is the edit target (#1349).
                     app.services.pasted = Some(pasted);
                     app.autosave = false;

apps/wordcraft/src/appimage_env.rs

@@ -0,0 +1,139 @@
+//! An AppImage's environment inherited from another program (#5653).
+//!
+//! The AppImage runtime gives the program it starts `APPIMAGE` (the image's file), `APPDIR` (where
+//! it is mounted), `ARGV0` and `OWD`, and whatever that program starts inherits them. So an AppImage
+//! app that opens a document through `xdg-open` starts an installed WordCraft with *its* `APPIMAGE`,
+//! and the self-update, which trusts that variable, replaces the other app's file with WordCraft.
+//!
+//! This copy is in the image only when its executable is under `APPDIR`. Otherwise the variables
+//! are another program's, and the process starts over without them, before anything reads them.
+//! (Removing them in place is `unsafe` under this edition, and `exec` keeps the process id, the
+//! arguments and the standard streams.)
+
+use std::ffi::OsStr;
+use std::io::Write;
+use std::os::unix::process::CommandExt;
+use std::path::Path;
+use std::process::Command;
+
+/// What the AppImage runtime sets; the same list as the documented workaround on the issue.
+const RUNTIME_VARS: [&str; 4] = ["APPIMAGE", "APPDIR", "ARGV0", "OWD"];
+
+/// Whether the executable runs from the image mounted at `appdir`. Anything unclear counts as not:
+/// the update then goes through the package or the plain file this copy really is, and at worst
+/// says it can't, where the other answer would overwrite a file that isn't WordCraft's.
+fn runs_from_image(exe: Option<&Path>, appdir: Option<&OsStr>) -> bool {
+    let (Some(exe), Some(appdir)) = (exe, appdir) else { return false };
+    let appdir = Path::new(appdir);
+    if !appdir.is_absolute() {
+        return false;
+    }
+    let Ok(appdir) = appdir.canonicalize() else { return false };
+    // No image mounts at the root, where every program on the machine would count as inside it.
+    appdir.parent().is_some() && exe.starts_with(&appdir)
+}
+
+/// Starts this process over without another program's AppImage variables; returns only when there
+/// are none to drop (`None`) or starting over failed. Call it first, before any other thread starts.
+pub(crate) fn drop_inherited() -> Option<std::io::Error> {
+    let appimage = std::env::var_os("APPIMAGE")?;
+    let exe = std::env::current_exe().ok();
+    if runs_from_image(exe.as_deref(), std::env::var_os("APPDIR").as_deref()) {
+        return None;
+    }
+    // The logger isn't up yet, and wouldn't survive `exec`. Unlike `eprintln!`, never panics.
+    let _ = writeln!(std::io::stderr(), "wordcraft: ignoring the AppImage {:?} inherited from another program", Path::new(&appimage));
+    let mut args = std::env::args_os();
+    // This file, whatever has happened to its path since it started.
+    let mut again = Command::new("/proc/self/exe");
+    if let Some(arg0) = args.next() {
+        again.arg0(arg0);
+    }
+    again.args(args);
+    for var in RUNTIME_VARS {
+        again.env_remove(var);
+    }
+    Some(again.exec())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::runs_from_image;
+    use std::path::{Path, PathBuf};
+
+    /// A folder of its own for one test, removed when it ends.
+    struct Tmp(PathBuf);
+
+    impl Tmp {
+        /// An empty folder named after `test` and this process, made afresh.
+        fn new(test: &str) -> Tmp {
+            let dir = std::env::temp_dir().join(format!("wordcraft-appimage-{test}-{}", std::process::id()));
+            let _ = std::fs::remove_dir_all(&dir);
+            std::fs::create_dir_all(&dir).unwrap();
+            Tmp(dir)
+        }
+    }
+
+    impl Drop for Tmp {
+        /// Removes the folder and what the test left in it.
+        fn drop(&mut self) {
+            let _ = std::fs::remove_dir_all(&self.0);
+        }
+    }
+
+    /// A mount with the executable inside it, by real path (`/tmp` may be a link).
+    fn mount(root: &Path, name: &str) -> (PathBuf, PathBuf) {
+        let dir = root.join(name);
+        std::fs::create_dir_all(dir.join("usr/bin")).unwrap();
+        let dir = dir.canonicalize().unwrap();
+        let exe = dir.join("usr/bin/wordcraft");
+        (dir, exe)
+    }
+
+    #[test]
+    fn an_executable_in_the_mount_is_the_image() {
+        let tmp = Tmp::new("own");
+        let (dir, exe) = mount(&tmp.0, ".mount_WordCr");
+        assert!(runs_from_image(Some(&exe), Some(dir.as_os_str())));
+    }
+
+    #[test]
+    fn a_link_to_the_mount_is_the_same_mount() {
+        let tmp = Tmp::new("link");
+        let (dir, exe) = mount(&tmp.0, ".mount_WordCr");
+        let link = tmp.0.join("link");
+        std::os::unix::fs::symlink(&dir, &link).unwrap();
+        assert!(runs_from_image(Some(&exe), Some(link.as_os_str())));
+    }
+
+    #[test]
+    fn an_installed_executable_is_not_in_another_programs_mount() {
+        let tmp = Tmp::new("other");
+        let (other, _) = mount(&tmp.0, ".mount_other");
+        assert!(!runs_from_image(Some(Path::new("/usr/bin/wordcraft")), Some(other.as_os_str())));
+    }
+
+    /// Paths compare by component: one mount's name can start with another's.
+    #[test]
+    fn a_mount_whose_name_starts_the_same_is_another_mount() {
+        let tmp = Tmp::new("prefix");
+        let (other, _) = mount(&tmp.0, ".mount_ab");
+        let (_, exe) = mount(&tmp.0, ".mount_abc");
+        assert!(!runs_from_image(Some(&exe), Some(other.as_os_str())));
+    }
+
+    #[test]
+    fn anything_unclear_is_not_the_image() {
+        let tmp = Tmp::new("unclear");
+        let (dir, exe) = mount(&tmp.0, ".mount_WordCr");
+        // No `APPDIR`, one that's gone (the other program has exited) or empty.
+        assert!(!runs_from_image(Some(&exe), None));
+        assert!(!runs_from_image(Some(&exe), Some(tmp.0.join("gone").as_os_str())));
+        assert!(!runs_from_image(Some(&exe), Some("".as_ref())));
+        // The root, and a relative path that happens to lead here.
+        assert!(!runs_from_image(Some(&exe), Some("/".as_ref())));
+        assert!(!runs_from_image(Some(&exe), Some(".".as_ref())));
+        // No executable path at all.
+        assert!(!runs_from_image(None, Some(dir.as_os_str())));
+    }
+}

apps/wordcraft/src/apple_events.rs

@@ -36,14 +36,15 @@ impl AppleEvents {
 
 /// Open the documents and act on the quit requests that arrived since the last frame. A document
 /// opens like File › Open: beside the open ones, in place of an untouched blank one, or, when it
-/// is already open, by showing it. Quitting closes the window, which asks about every document
+/// is already open, by showing it; a Word template starts a new document based on it, as Word's
+/// default verb for one does (#3433). Quitting closes the window, which asks about every document
 /// with unsaved changes.
 pub fn poll(inbox: &Inbox, app: &mut WordApp, ctx: &egui::Context) {
     for e in inbox.drain() {
         match e {
             Event::Open(paths) => {
                 for p in paths.iter().filter_map(|p| p.to_str()) {
-                    match app.open_file(p, None) {
+                    match app.open_from_platform(p, None) {
                         // Leave Backstage, as opening from the Open dialog does.
                         Ok(_) => app.ui.backstage = false,
                         Err(e) => {

apps/wordcraft/src/main.rs

@@ -10,6 +10,8 @@
 #![deny(clippy::unwrap_used, clippy::expect_used, clippy::panic, clippy::unimplemented, clippy::todo, clippy::unreachable)]
 
 mod animations;
+#[cfg(target_os = "linux")]
+mod appimage_env;
 #[cfg(target_os = "macos")]
 mod apple_events;
 mod clipboard;
@@ -67,6 +69,8 @@ impl eframe::App for App {
         if let Some((pos, size)) = self.1.and_then(|g| g.fit_on_monitor(&window)) {
             ctx.send_viewport_cmd(egui::ViewportCommand::InnerSize(size));
             ctx.send_viewport_cmd(egui::ViewportCommand::OuterPosition(pos));
+            let b = self.0.window_builder().clone().with_inner_size(size);
+            self.0.set_window_builder(b);
         }
         // The geometry saved for the next start is the active window's.
         let prev = self.0.ui.window;
@@ -421,14 +425,15 @@ fn without_launch_args(args: Vec<String>) -> Vec<String> {
 
 /// Open a file whose name isn't valid Unicode (#1440). `file.open` takes its path as text, so it
 /// gets the file's bytes under the lossy name, and the document then keeps the real path to save to.
-/// The bytes are read as `file.open` reads a path: only a regular file, and at most 2 GB.
+/// The bytes are read as `file.open` reads a path: only a regular file, and at most 2 GB. A Word
+/// template starts a new document instead ([`WordApp::open_from_platform`]), which has no path.
 fn open_non_unicode(app: &mut WordApp, path: &std::path::Path) {
     let name = path.to_string_lossy();
     let opened = wordcraft_engine::io::read_document(path)
-        .and_then(|bytes| app.run("file.open", serde_json::json!({"path": name, "data": wordcraft_engine::cmd::insert::base64_encode(&bytes)})));
+        .and_then(|bytes| app.open_from_platform(&name, Some(&wordcraft_engine::cmd::insert::base64_encode(&bytes))));
     match opened {
         // Not when it waits for a password: that open names the document itself.
-        Ok(r) if r.get("pending").is_none() => app.session.path = Some(path.into()),
+        Ok(r) if r.get("pending").is_none() && !wordcraft_ui_egui::documents::is_word_template(&name) => app.session.path = Some(path.into()),
         Ok(_) => {}
         Err(e) => {
             app.report_open_failure(&name, &e);
@@ -449,9 +454,22 @@ fn main() -> eframe::Result {
     if let Some(result) = craft_update::self_update::finish(&wordcraft_ui_egui::UPDATE_APP, &std::env::args_os().skip(1).collect::<Vec<_>>()) {
         std::process::exit(i32::from(result.is_err()));
     }
+    // Linux: before anything reads another program's AppImage variables, or a thread starts (#5653).
+    #[cfg(target_os = "linux")]
+    let start_over_failed = appimage_env::drop_inherited();
+    // Still carrying them, this process must never update itself: the update would replace the
+    // other program's file.
+    #[cfg(target_os = "linux")]
+    let updates_unsafe = start_over_failed.is_some();
+    #[cfg(not(target_os = "linux"))]
+    let updates_unsafe = false;
     // First, so the panic hook and every start-up warning are recorded (`logging`).
     let logger = logging::install();
     logging::install_panic_hook();
+    #[cfg(target_os = "linux")]
+    if let Some(e) = start_over_failed {
+        log::error!("couldn't start again without another program's AppImage variables, so the self-update is off in this window: {e}");
+    }
     let mut control_port: Option<u16> = std::env::var("WORDCRAFT_CONTROL_PORT").ok().and_then(|p| p.parse().ok());
     let mut files = Vec::new();
     let mut sample = false;
@@ -566,6 +584,10 @@ fn main() -> eframe::Result {
             let mut app = WordApp::new(session, services(file_dialog, &cc.egui_ctx));
             app.set_startup_update_check(true);
             load_prefs(&mut app);
+            // After the preferences, so the reason it gives is in the interface language.
+            if updates_unsafe {
+                app.disable_self_update();
+            }
             load_user_dictionary(&mut app);
             // Language packs from the suite's font store; an Install click is the person's own.
             wordcraft_engine::packs::use_font_store(true);
@@ -605,8 +627,9 @@ fn main() -> eframe::Result {
             if files.is_empty() && other_files.is_empty() && !sample && control_port.is_none() {
                 app.start_screen();
             }
+            // A Word template named here starts a new document based on it (#3433).
             for f in files {
-                if let Err(e) = app.run("file.open", serde_json::json!({"path": f})) {
+                if let Err(e) = app.open_from_platform(&f, None) {
                     app.report_open_failure(&f, &e);
                 }
             }

apps/wordcraft/tests/it/appimage_env.rs

@@ -0,0 +1,99 @@
+//! A WordCraft started by an AppImage app inherits that app's `APPIMAGE`, which the self-update
+//! would take for this copy's file and replace (#5653). The real program, started both ways.
+
+use std::ffi::OsString;
+use std::os::unix::ffi::OsStringExt;
+use std::path::{Path, PathBuf};
+use std::process::{Command, Output, Stdio};
+use std::time::{Duration, Instant};
+
+const NOTICE: &str = "ignoring the AppImage";
+
+/// A folder of its own for one test, removed when it ends.
+struct Tmp(PathBuf);
+
+impl Tmp {
+    /// An empty folder named after `test` and this process, made afresh.
+    fn new(test: &str) -> Tmp {
+        let dir = std::env::temp_dir().join(format!("wordcraft-appimage-it-{test}-{}", std::process::id()));
+        let _ = std::fs::remove_dir_all(&dir);
+        std::fs::create_dir_all(&dir).unwrap();
+        Tmp(dir)
+    }
+}
+
+impl Drop for Tmp {
+    /// Removes the folder and what the test left in it.
+    fn drop(&mut self) {
+        let _ = std::fs::remove_dir_all(&self.0);
+    }
+}
+
+/// Runs the real app with `args`, as an AppImage app with `appimage` mounted at `appdir` would
+/// start it, and waits up to 30 s for it to exit (with no display it can't open a window).
+fn app(args: &[OsString], appimage: &Path, appdir: &Path) -> Output {
+    let mut child = Command::new(env!("CARGO_BIN_EXE_wordcraft"))
+        .args(args)
+        .env("APPIMAGE", appimage)
+        .env("APPDIR", appdir)
+        .env("ARGV0", "other-app")
+        .env("OWD", "/")
+        // With no display, an app that tries to open a window fails at once instead.
+        .env_remove("DISPLAY")
+        .env_remove("WAYLAND_DISPLAY")
+        .env_remove("WAYLAND_SOCKET")
+        .env_remove("WORDCRAFT_CONTROL_PORT")
+        .env("WORDCRAFT_NO_PREFS", "1")
+        .stdin(Stdio::null())
+        .stdout(Stdio::piped())
+        .stderr(Stdio::piped())
+        .spawn()
+        .unwrap();
+    // A start-over that kept the variables would start over for ever, and end here.
+    let deadline = Instant::now() + Duration::from_secs(30);
+    while child.try_wait().unwrap().is_none() {
+        if Instant::now() > deadline {
+            let _ = child.kill();
+            panic!("wordcraft {args:?} was still running after 30 s");
+        }
+        std::thread::sleep(Duration::from_millis(50));
+    }
+    child.wait_with_output().unwrap()
+}
+
+/// The notice comes once: the program it started over as no longer sees the variables.
+#[test]
+fn another_programs_variables_are_dropped_once_and_the_program_carries_on() {
+    let tmp = Tmp::new("dropped");
+    let other = tmp.0.join("other-app.AppImage");
+    std::fs::write(&other, b"another program").unwrap();
+    let out = app(&["--version".into()], &other, &tmp.0);
+    let (stdout, stderr) = (String::from_utf8_lossy(&out.stdout), String::from_utf8_lossy(&out.stderr));
+    assert!(out.status.success(), "status {:?}, stderr: {stderr}", out.status);
+    assert!(stdout.starts_with("wordcraft "), "stdout: {stdout}");
+    assert_eq!(stderr.matches(NOTICE).count(), 1, "stderr: {stderr}");
+    assert!(stderr.contains("other-app.AppImage"), "stderr: {stderr}");
+    assert_eq!(std::fs::read(&other).unwrap(), b"another program");
+}
+
+/// Arguments reach the program it starts over as byte for byte, even one that isn't Unicode.
+#[test]
+fn starting_over_keeps_the_arguments() {
+    let tmp = Tmp::new("args");
+    let out = app(&["--sample".into(), "--control".into(), OsString::from_vec(b"79\xe981".to_vec())], &tmp.0.join("gone"), &tmp.0);
+    let stderr = String::from_utf8_lossy(&out.stderr);
+    assert_eq!(out.status.code(), Some(2), "stderr: {stderr}");
+    assert!(stderr.contains("argument 3 is not valid Unicode: 79\u{FFFD}81"), "stderr: {stderr}");
+}
+
+/// Its own image's variables stay, or the self-update couldn't find the file to replace.
+#[test]
+fn its_own_images_variables_stay() {
+    let exe = Path::new(env!("CARGO_BIN_EXE_wordcraft")).canonicalize().unwrap();
+    let mount = exe.parent().unwrap();
+    let out = app(&["--version".into()], &mount.join("WordCraft.AppImage"), mount);
+    let (stdout, stderr) = (String::from_utf8_lossy(&out.stdout), String::from_utf8_lossy(&out.stderr));
+    assert!(out.status.success(), "status {:?}, stderr: {stderr}", out.status);
+    assert!(stdout.starts_with("wordcraft "), "stdout: {stdout}");
+    assert!(!stderr.contains(NOTICE), "stderr: {stderr}");
+}

apps/wordcraft/tests/it/main.rs

@@ -1,3 +1,5 @@
+#[cfg(target_os = "linux")]
+mod appimage_env;
 mod control_arg;
 mod help;
 #[cfg(unix)]

contributors/people.toml

@@ -14,6 +14,10 @@
 # To remove your names, delete your entry (or set fields empty) in a PR. The next
 # contributors.py run, and the next build, drop them.
 
+[people.SenuraMalaka]
+real_name = "Senura Dissanayake"
+sync_github_name = true
+
 [people.divyanshu2074]
 real_name = "Divyanshu Gupta"
 sync_github_name = true
@@ -197,3 +201,6 @@ display_name = "entropybeater"
 
 [people.helenkwok]
 real_name = "Helen Kwok"
+
+[people.ashokram-si]
+real_name = "Ashok Ram"

crates/doc/src/cfb_guard.rs

@@ -7,8 +7,9 @@
 /// makes it allocate gigabytes. Here the DIFAT is walked the way `cfb` walks it (header entries,
 /// then the DIFAT chain, at most one step per sector; trailing zero and free entries trimmed
 /// the same way) and the file is refused when it lists one FAT sector twice, or more FAT sectors
-/// than its sectors need (one FAT sector covers `sector_len / 4` of them, plus one spare), so the
-/// FAT `cfb` builds is about the size the file needs. Anything else that is wrong is left for
+/// than the file has sectors — every FAT sector is a sector of the file, so a longer list is
+/// duplicates or names `cfb` refuses anyway, and the FAT `cfb` builds (one sector of entries per
+/// listed name) stays within about the file's own size. Anything else that is wrong is left for
 /// `cfb` to report.
 pub fn check(bytes: &[u8]) -> Result<(), String> {
     const MAGIC: [u8; 8] = [0xD0, 0xCF, 0x11, 0xE0, 0xA1, 0xB1, 0x1A, 0xE1];
@@ -52,12 +53,14 @@ pub fn check(bytes: &[u8]) -> Result<(), String> {
     while difat.last() == Some(&FREE) {
         difat.pop();
     }
-    // Each listed FAT sector adds a whole sector of entries. A FAT sector holds the entries of
-    // `sector_len / 4` sectors, so a file needs no more FAT sectors than that covers (one spare),
-    // and none of them twice. A free entry in the middle is left for `cfb` to refuse.
+    // Each listed FAT sector adds a whole sector of entries. A real file lists no more FAT
+    // sectors than the file has — every FAT sector is a sector of the file, and none of them
+    // twice (checked below) — so the FAT `cfb` builds stays within about the file's own size.
+    // The #3642 blowup listed 2048 × 127 names in a ~1 MB file. A tighter bound (one FAT sector
+    // per `sector_len / 4` sectors plus a spare) refused real files, whose writers reserve FAT
+    // for growth: an encrypted package of 78 sectors lists 4 and didn't open (#5109).
     let listed: Vec<u32> = difat.into_iter().filter(|&s| s != FREE).collect();
-    let needed = num_sectors.div_ceil(sector_len / 4).saturating_add(1);
-    if listed.len() > needed {
+    if listed.len() > num_sectors {
         return Err(format!("the DIFAT lists {} FAT sectors in a file of {num_sectors} sectors", listed.len()));
     }
     let mut seen = std::collections::HashSet::with_capacity(listed.len());
@@ -123,10 +126,16 @@ mod tests {
     }
 
     #[test]
-    fn a_fat_sector_listed_twice_or_more_than_the_file_needs_is_refused() {
-        // Each list is shorter than the file has sectors, which the guard used to accept.
+    fn a_fat_sector_listed_twice_or_more_than_the_file_has_is_refused() {
+        // A duplicate, and a list longer than the file has sectors: every FAT sector is a
+        // sector of the file, so a longer list is one of the two.
         assert!(super::check(&header_difat(10, &[0, 0])).is_err_and(|e| e.contains("FAT sector 0 twice")));
-        assert!(super::check(&header_difat(10, &[0, 1, 2])).is_err_and(|e| e.contains("3 FAT sectors")));
+        assert!(
+            super::check(&header_difat(10, &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10])).is_err_and(|e| e.contains("11 FAT sectors in a file of 10 sectors"))
+        );
+        // A writer may reserve FAT for growth: a real encrypted package of 78 sectors lists
+        // 4 FAT sectors and didn't open since the #3642 guard landed (#5109).
+        assert!(super::check(&header_difat(78, &[0, 1, 2, 3])).is_ok());
         assert!(super::check(&header_difat(10, &[0])).is_ok());
         assert!(super::check(&header_difat(200, &[0, 1])).is_ok());
     }

crates/doc/src/edit.rs

@@ -772,6 +772,10 @@ impl Document {
         if a.path.parent() == b.path.parent() {
             // Same container: trim the ends, drop the blocks between, join.
             let (ia, ib) = (a.path.last(), b.path.last());
+            // Removing an already-empty paragraph must not impose its style or mark formatting
+            // on the following content. Test before trimming: deleting actual heading text is
+            // still an ordinary merge that keeps the first paragraph's properties.
+            let empty_start = self.para(a.story, &a.path).is_some_and(Paragraph::is_empty);
             // Kept markers of the paragraphs in between move to the join.
             let mut moved: Vec<(InlineObject, CharProps)> = Vec::new();
             // `kept` is sorted: one walk of each paragraph's objects and runs (#3614).
@@ -819,15 +823,18 @@ impl Document {
             if ib < bl.len() && ia < ib {
                 bl.drain(ia + 1..=ib);
             }
-            tail.props = ParaProps::default();
             // The deleted marks include `a`'s, so a section break there goes too: the joined
             // paragraph ends where `b`'s mark did, and its text takes the following section's
             // setup, as in Word (#2127).
             let sect = tail.section.take();
             let pa = self.para_mut(a.story, &a.path)?;
-            let keep_props = pa.props.clone();
+            let keep_props = if empty_start { tail.props.clone() } else { pa.props.clone() };
+            let keep_mark = empty_start.then(|| tail.mark.clone());
             pa.append(tail);
             pa.props = keep_props;
+            if let Some(mark) = keep_mark {
+                pa.mark = mark;
+            }
             let before = std::mem::replace(&mut pa.section, sect);
             if a.story == crate::StoryRef::Body && a.path.0.len() == 1 {
                 // Nearest first, as "link to previous" resolved them before the delete.

crates/doc/src/kept.rs

@@ -478,8 +478,9 @@ fn family_fingerprint(doc: &crate::Document, root: u32, replies: &HashMap<u32, V
 }
 
 /// What `doc`'s page setup is: each section's setup (where it starts, its size, margins,
-/// columns, headers and footers), the headers' and footers' content, the page colour and
-/// picture, and the notes' numbering.
+/// columns, headers and footers), the headers' and footers' content, whether odd and even pages
+/// differ, the page colour and picture, the watermark (one with no text isn't written), and the
+/// notes' numbering.
 pub fn pages_fingerprint(doc: &crate::Document) -> u64 {
     let mut sections: Vec<&crate::SectionProps> = doc.body.iter().filter_map(|b| b.as_para()).filter_map(|p| p.section.as_deref()).collect();
     sections.push(&doc.last_section);
@@ -490,8 +491,10 @@ pub fn pages_fingerprint(doc: &crate::Document) -> u64 {
     match serde_json::to_vec(&(
         sections,
         parts,
+        s.even_odd_headers,
         &s.page_color,
         &s.page_picture,
+        s.watermark.as_ref().filter(|w| !w.text.trim().is_empty()),
         (&s.footnote_format, &s.endnote_format, &s.footnote_pr, &s.endnote_pr),
         (&doc.footnote_separators, &doc.endnote_separators),
     )) {

crates/doc/src/link.rs

@@ -125,6 +125,7 @@ mod tests {
             ("HYPERLINK\t\"http://example.com\"", Some("http://example.com")),
             (r#"HYPERLINK"http://example.com""#, Some("http://example.com")),
             (r#"HYPERLINK "http://example.com/path\latest""#, Some("http://example.com/pathlatest")),
+            (r#"HYPERLINK "x \* Upper""#, Some("x * Upper")),
             (r#"HYPERLINK "C:\\Docs\\My file.docx""#, Some(r"C:\Docs\My file.docx")),
             (r#"HYPERLINK "say \"hi\"""#, Some(r#"say "hi""#)),
             (r#"HYPERLINK " http://example.com ""#, Some("http://example.com")),

crates/doc/src/numbering.rs

@@ -228,7 +228,7 @@ impl Numbering {
     /// Whether a paragraph in list `num` at `level` gets a label ([`Counters::next_label`] gives
     /// one), without counting it.
     pub fn shows_label(&self, num: u32, level: u8) -> bool {
-        self.abstract_of(num).is_some() && self.level(num, level.min(8)).is_some_and(|d| d.format != NumFormat::None)
+        self.abstract_of(num).is_some() && self.level(num, level.min(8)).is_some()
     }
     pub fn level(&self, num: u32, level: u8) -> Option<&Level> {
         if let Some((_, l)) = self.num(num)?.level_overrides.iter().find(|(i, _)| *i == level) {
@@ -325,12 +325,9 @@ impl Counters {
         let lv = (level as usize).min(8);
         // This list's own definition of a level wins over its abstract list's.
         let level_def = |k: usize| numbering.level(num, k as u8);
+        // A "none" level (Word's numFmt none) still counts and restarts the levels below it; only
+        // its number reads as nothing, so its literal text ("RESPONSE:") shows (#4985).
         let def = level_def(lv)?.clone();
-        // A "none" level shows no label at all (Word's numFmt none) — not even its text.
-        if def.format == NumFormat::None {
-            self.labels.insert((abs.id, lv as u8), CountedLabel { label: None, levels: Vec::new() });
-            return None;
-        }
         let st = self.state.entry(abs.id).or_insert([0; 9]);
         let fl = self.flags.entry(abs.id).or_default();
         // Only this level's override, not the num's others: the counters are shared by every
@@ -382,7 +379,14 @@ impl Counters {
         let started = *fl;
         let numeric = def.format != NumFormat::Bullet;
         let label = if numeric { render_label(numbering, num, &def, &vals, &started) } else { def.text.clone() };
-        self.labels.insert((abs.id, lv as u8), CountedLabel { label: numeric.then(|| label.clone()), levels: template_levels(&def.text) });
+        // A "none" level shows its literal text, but has no number of its own for a reference to
+        // show (#4985): it is kept as one that shows none.
+        let counted = if def.format == NumFormat::None {
+            CountedLabel { label: None, levels: Vec::new() }
+        } else {
+            CountedLabel { label: numeric.then(|| label.clone()), levels: template_levels(&def.text) }
+        };
+        self.labels.insert((abs.id, lv as u8), counted);
         Some((label, def))
     }
     /// Which levels of `num`'s list have counted a paragraph yet (a level never used has no
@@ -455,9 +459,9 @@ fn render_label(numbering: &Numbering, num: u32, def: &Level, vals: &[u32; 9], s
             let counted = started.get(k).is_some_and(|f| f.started);
             let val = if counted { val } else { level_def(k).map(|l| l.start).unwrap_or(1) };
             // Legal numbering (w:isLgl) shows every level in Arabic digits, whatever its own
-            // format (二 of a Chinese outline becomes 2 in "2.1").
+            // format (二 of a Chinese outline becomes 2 in "2.1"); a "none" level stays empty.
             let fmt = match level_def(k).map(|l| l.format).unwrap_or(NumFormat::Decimal) {
-                NumFormat::DecimalZero => NumFormat::DecimalZero,
+                f @ (NumFormat::DecimalZero | NumFormat::None) => f,
                 _ if def.legal => NumFormat::Decimal,
                 f => f,
             };
@@ -556,6 +560,27 @@ mod tests {
         assert_eq!(n.level(id, 2).map(|l| l.format), Some(NumFormat::DecimalZero));
     }
 
+    #[test]
+    fn a_none_level_shows_its_text_and_still_counts() {
+        // As Word shows them (#4985): a "none" level's literal text is its label, its own number
+        // reads as nothing (also in legal numbering), and it restarts the levels below it.
+        let mut n = Numbering::default();
+        let id = n.add_list(ListKind::Numbered);
+        let lvl = |format, text: &str| Level { format, text: text.into(), ..Level::default() };
+        let legal = Level { legal: true, ..lvl(NumFormat::Decimal, "%2.%3") };
+        if let Some(num) = n.nums.iter_mut().find(|x| x.id == id) {
+            num.level_overrides = vec![(0, lvl(NumFormat::None, "RESPONSE:")), (1, lvl(NumFormat::None, "[%2]")), (2, legal)];
+        }
+        assert!(n.shows_label(id, 0));
+        let mut c = Counters::default();
+        assert_eq!(c.next_label(&n, id, 0).unwrap().0, "RESPONSE:");
+        assert_eq!(c.next_label(&n, id, 1).unwrap().0, "[]");
+        assert_eq!(c.next_label(&n, id, 2).unwrap().0, ".1");
+        assert_eq!(c.next_label(&n, id, 2).unwrap().0, ".2");
+        assert_eq!(c.next_label(&n, id, 1).unwrap().0, "[]");
+        assert_eq!(c.next_label(&n, id, 2).unwrap().0, ".1");
+    }
+
     #[test]
     fn a_start_override_restarts_only_its_own_level_of_a_shared_list() {
         // One abstract list numbers the headings through two nums: one for levels 1-4, and one

crates/doc/src/para/float.rs

@@ -186,6 +186,11 @@ pub struct Float {
     /// top. 0 = unknown, drawn in document order (objects inserted here, or from other formats).
     #[serde(skip_serializing_if = "is_zero_u32")]
     pub z: u32,
+    /// Turned, but the file keeps no room for the turn (`wp:effectExtent` missing or zero, as
+    /// files from other tools have): Word then keeps an inline one's line to its frame and lets
+    /// the turned object overhang it (see [`Float::line_pad`]). Cleared by [`Float::set_spin`].
+    #[serde(skip_serializing_if = "is_false")]
+    pub bare_turn: bool,
 }
 
 /// Longest alternative text kept (characters): hostile files and commands are capped.
@@ -241,6 +246,7 @@ impl Float {
         self.rot = wordcraft_geom::normalize_degrees(s.deg);
         self.flip_h = s.flip_h;
         self.flip_v = s.flip_v;
+        self.bare_turn = false;
     }
     /// How far a `w` × `h` frame's rotated bounds reach past it on each side (x, y): the room a
     /// rotated object takes beyond its frame (zero unrotated; negative where the turned object
@@ -250,6 +256,23 @@ impl Float {
         let (bw, bh) = self.spin().extent(w, h);
         ((bw - w) / 2.0, (bh - h) / 2.0)
     }
+    /// A [`Float::bare_turn`] still inline and without effects room: made floating, or given room
+    /// for a shadow or glow since, it saves, and so lays out, with the turn's room again.
+    pub fn is_bare_turn(&self) -> bool {
+        self.bare_turn && self.wrap == Wrap::Inline && self.effect_extent() == [0.0; 4]
+    }
+    /// The room an inline `w` × `h` object's line keeps past its frame on each side (x, y) for its
+    /// turn: its rotated bounds ([`Float::spin_pad`]) — or, for a [`Float::is_bare_turn`], none, except
+    /// that Word 2013+ layout (`modern`) exchanges the frame's width and height from 45° up to 135°
+    /// and from 225° up to 315°.
+    pub fn line_pad(&self, w: f32, h: f32, modern: bool) -> (f32, f32) {
+        if !self.is_bare_turn() {
+            return self.spin_pad(w, h);
+        }
+        let (w, h) = (wordcraft_geom::finite(w).clamp(0.0, 4000.0), wordcraft_geom::finite(h).clamp(0.0, 4000.0));
+        let deg = wordcraft_geom::normalize_degrees(self.rot);
+        if modern && ((45.0..135.0).contains(&deg) || (225.0..315.0).contains(&deg)) { ((h - w) / 2.0, (w - h) / 2.0) } else { (0.0, 0.0) }
+    }
     /// The effect extents, finite and clamped (left, top, right, bottom).
     pub fn effect_extent(&self) -> [f32; 4] {
         self.effect.map(|v| if v.is_finite() { v.clamp(0.0, 1584.0) } else { 0.0 })

crates/doc/src/props.rs

@@ -27,7 +27,8 @@ pub use char_props::{
 };
 pub use color::{Rgb, TextColor};
 pub use para_props::{
-    Align, Border, BorderStyle, Borders, LineSpacing, NumRef, ParaProps, TAB_POS_EPS, TabAlign, TabLeader, TabStop, TextAlignment, TextFrame,
+    Align, Border, BorderStyle, Borders, LineSpacing, NumRef, ParaProps, ParagraphTextDirection, TAB_POS_EPS, TabAlign, TabLeader, TabStop,
+    TextAlignment, TextFrame,
 };
 pub use table_props::{
     CellPicture, CellProps, HeightRule, PictureRepeat, RowProps, TableFloat, TableLook, TableProps, TextDirection, TextVert, VAlign, VMerge,

crates/doc/src/props/para_props.rs

@@ -26,6 +26,67 @@ impl Align {
     }
 }
 
+/// Paragraph-level text flow (`w:pPr/w:textDirection`, ECMA-376 §17.3.1.41).
+///
+/// Kept separate from [`super::TextDirection`], whose three values model the table-cell direction
+/// control. These values preserve the exact OOXML direction through JSON and DOCX saves.
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub enum ParagraphTextDirection {
+    LrTb,
+    LrTbV,
+    Tb,
+    TbRl,
+    TbRlV,
+    TbLr,
+    TbLrV,
+    BtLr,
+    TbV,
+    Rl,
+    RlV,
+    Lr,
+    LrV,
+}
+
+impl ParagraphTextDirection {
+    pub fn from_ooxml(value: &str) -> Option<Self> {
+        match value {
+            "lrTb" => Some(Self::LrTb),
+            "lrTbV" => Some(Self::LrTbV),
+            "tb" => Some(Self::Tb),
+            "tbRl" => Some(Self::TbRl),
+            "tbRlV" => Some(Self::TbRlV),
+            "tbLr" => Some(Self::TbLr),
+            "tbLrV" => Some(Self::TbLrV),
+            "btLr" => Some(Self::BtLr),
+            "tbV" => Some(Self::TbV),
+            "rl" => Some(Self::Rl),
+            "rlV" => Some(Self::RlV),
+            "lr" => Some(Self::Lr),
+            "lrV" => Some(Self::LrV),
+            _ => None,
+        }
+    }
+
+    pub fn ooxml(self) -> &'static str {
+        match self {
+            Self::LrTb => "lrTb",
+            Self::LrTbV => "lrTbV",
+            Self::Tb => "tb",
+            Self::TbRl => "tbRl",
+            Self::TbRlV => "tbRlV",
+            Self::TbLr => "tbLr",
+            Self::TbLrV => "tbLrV",
+            Self::BtLr => "btLr",
+            Self::TbV => "tbV",
+            Self::Rl => "rl",
+            Self::RlV => "rlV",
+            Self::Lr => "lr",
+            Self::LrV => "lrV",
+        }
+    }
+}
+
 /// Vertical alignment of characters of different sizes on a line (`w:textAlignment`,
 /// ECMA-376 §17.3.1.39).
 #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize)]
@@ -391,6 +452,9 @@ pub struct ParaProps {
     pub suppress_line_numbers: Option<bool>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub bidi: Option<bool>,
+    /// Paragraph text direction (`w:textDirection`); `None` inherits from the paragraph style.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub text_direction: Option<ParagraphTextDirection>,
     /// Drop cap: lines to drop (0 = none) — applies to the first character(s).
     #[serde(skip_serializing_if = "Option::is_none")]
     pub drop_cap: Option<u8>,
@@ -462,6 +526,7 @@ impl ParaProps {
     }
 
     fn overlay_script_layout(&mut self, patch: &ParaProps) {
+        overlay_fields!(self, patch; bidi, text_direction, kinsoku, word_wrap, overflow_punct, top_line_punct, auto_space_de, auto_space_dn);
         overlay_fields!(self, patch; bidi, kinsoku, word_wrap, overflow_punct, top_line_punct, auto_space_de, auto_space_dn, text_alignment);
     }
     pub fn overlaid(mut self, patch: &ParaProps) -> ParaProps {

crates/doc/src/resolve.rs

@@ -137,9 +137,12 @@ impl ResolvedChar {
     pub fn uses_complex(&self, c: char) -> bool {
         self.rtl || self.cs || crate::bidi::is_complex_script(c)
     }
-    /// The size glyphs are drawn at (super/subscript shrink to ~2/3).
+    /// The size glyphs are drawn and measured at. Word sets superscript and subscript text at
+    /// 0.65 of the run's size rounded to the half point, a font size the format can store
+    /// (measured in Word for Mac from digit advances: 8 pt → 5, 9 → 6, 11 → 7, 12 → 8,
+    /// 16 → 10.5, 24 → 15.5, 36 → 23.5).
     pub fn draw_size(&self) -> f32 {
-        if self.vert_align == VertAlign::Baseline { self.size } else { self.size * 0.65 }
+        if self.vert_align == VertAlign::Baseline { self.size } else { (self.size * 0.65 * 2.0).round() / 2.0 }
     }
     /// Baseline shift, points (positive = up).
     pub fn baseline_shift(&self) -> f32 {
@@ -453,6 +456,32 @@ fn diff_patch<T: Serialize + serde::de::DeserializeOwned + Default>(with: &T, wi
 mod tests {
     use super::*;
 
+    /// Superscript and subscript sizes as Word for Mac sets them (measured from digit advances in
+    /// its PDF export): 0.65 of the size, rounded to the half point.
+    #[test]
+    fn raised_and_lowered_text_is_sized_to_the_half_point() {
+        for (size, want) in [
+            (8.0, 5.0),
+            (9.0, 6.0),
+            (10.0, 6.5),
+            (11.0, 7.0),
+            (12.0, 8.0),
+            (13.0, 8.5),
+            (14.0, 9.0),
+            (16.0, 10.5),
+            (20.0, 13.0),
+            (24.0, 15.5),
+            (36.0, 23.5),
+        ] {
+            for vert_align in [VertAlign::Superscript, VertAlign::Subscript] {
+                let rc = ResolvedChar { size, vert_align, ..StyleSheet::builtin().resolve_char(None, &CharProps::default()) };
+                assert_eq!(rc.draw_size(), want, "{size} pt {vert_align:?}");
+            }
+        }
+        let rc = StyleSheet::builtin().resolve_char(None, &CharProps { size: Some(11.0), ..Default::default() });
+        assert_eq!(rc.draw_size(), 11.0, "baseline text keeps its size");
+    }
+
     /// #2036: only a character itself (direct formatting) has a symbol font of its own; a
     /// style never gives one to its text.
     #[test]

crates/doc/src/section.rs

@@ -289,13 +289,14 @@ impl NumFormat {
     pub fn from_ooxml(s: &str) -> NumFormat {
         NumFormat::ALL.into_iter().find(|f| f.ooxml() == s).unwrap_or(NumFormat::Decimal)
     }
-    /// Format `n` (1-based). Formats WordCraft doesn't draw yet (Hebrew, Hindi, Thai,
-    /// Vietnamese, text amounts) show decimal, as do values past a sequence's end
-    /// (the 21st item of ①…⑳).
+    /// Format `n` (1-based). Formats WordCraft doesn't draw yet (Hebrew, Hindi, Thai letters
+    /// and counting, Vietnamese, text amounts) show decimal, as do values past a sequence's end
+    /// (the 21st item of ①…⑳). Thai numbers are the decimal number in Thai digits (๑, ๒ … ๑๐).
     pub fn format(self, n: u32) -> String {
         use crate::cjk_numbers as cjk;
         match self {
             NumFormat::DecimalZero => format!("{n:02}"),
+            NumFormat::ThaiNumbers => thai_digits(n),
             NumFormat::UpperRoman => roman(n).to_uppercase(),
             NumFormat::LowerRoman => roman(n),
             NumFormat::UpperLetter => letters(n).to_uppercase(),
@@ -349,7 +350,6 @@ impl NumFormat {
             | NumFormat::HindiNumbers
             | NumFormat::HindiCounting
             | NumFormat::ThaiLetters
-            | NumFormat::ThaiNumbers
             | NumFormat::ThaiCounting
             | NumFormat::BahtText
             | NumFormat::DollarText => n.to_string(),
@@ -357,6 +357,11 @@ impl NumFormat {
     }
 }
 
+/// `n` in Thai digits, ๐ (U+0E50) to ๙ (#5757).
+fn thai_digits(n: u32) -> String {
+    n.to_string().chars().filter_map(|c| c.to_digit(10).and_then(|d| char::from_u32(0x0E50 + d))).collect()
+}
+
 /// Chicago Manual of Style note marks: *, †, ‡, §, then doubled (**, ††, …) and so on.
 fn chicago(n: u32) -> String {
     const MARKS: [char; 4] = ['*', '\u{2020}', '\u{2021}', '\u{a7}'];
@@ -999,6 +1004,18 @@ mod tests {
         assert_eq!(NumFormat::from_ooxml("klingon"), NumFormat::Decimal, "unknown formats are decimal");
     }
 
+    /// #5757: Thai numbers are the decimal number in Thai digits, every digit, at any length.
+    #[test]
+    fn thai_numbers_use_thai_digits() {
+        let labels: Vec<String> = [1, 2, 3, 9, 10, 11, 20, 105, 2569].iter().map(|n| NumFormat::ThaiNumbers.format(*n)).collect();
+        assert_eq!(labels.join(" "), "๑ ๒ ๓ ๙ ๑๐ ๑๑ ๒๐ ๑๐๕ ๒๕๖๙");
+        assert_eq!(NumFormat::ThaiNumbers.format(0), "๐");
+        assert_eq!(NumFormat::ThaiNumbers.format(u32::MAX), "๔๒๙๔๙๖๗๒๙๕");
+        assert_eq!(NumFormat::from_ooxml("thaiNumbers"), NumFormat::ThaiNumbers);
+        // Formats not drawn yet still show decimal.
+        assert_eq!(NumFormat::ThaiLetters.format(3), "3");
+    }
+
     #[test]
     fn russian_letters_follow_word() {
         // Word's own labels for a russianLower list: no ё, й, ъ or ь, then doubled letters.

crates/doc/src/symbol_fonts.rs

@@ -19,8 +19,8 @@
 //!
 //! Tables are indexed by code − 0x20; 0 = no Unicode equivalent.
 
-/// The symbol fonts this module knows, by lowercase family name.
-fn table(family: &str) -> Option<&'static [u32; 224]> {
+/// The symbol table for a single family name, by lowercase family name.
+fn single_table(family: &str) -> Option<&'static [u32; 224]> {
     let f = family.trim();
     if f.eq_ignore_ascii_case("symbol") || f.eq_ignore_ascii_case("symbol (pcl6)") {
         Some(&SYMBOL)
@@ -37,6 +37,11 @@ fn table(family: &str) -> Option<&'static [u32; 224]> {
     }
 }
 
+/// The symbol fonts this module knows, by lowercase family name.
+fn table(family: &str) -> Option<&'static [u32; 224]> {
+    if family.contains(';') { family.split(';').map(str::trim).filter(|s| !s.is_empty()).find_map(single_table) } else { single_table(family) }
+}
+
 /// Whether `family` is one of Word's symbol fonts (Symbol, Wingdings, Wingdings 2, Wingdings 3,
 /// Webdings), whose characters [`symbol_char`] maps.
 pub fn is_symbol_font(family: &str) -> bool {

crates/doc/tests/empty_paragraph_merge.rs

@@ -0,0 +1,158 @@
+use std::sync::Arc;
+use wordcraft_doc::props::{Align, CharProps, ParaProps};
+use wordcraft_doc::{Block, Document, Paragraph, Path, Pos, StoryRef, Table, para_block};
+
+/// Build an empty heading followed by explicitly formatted, mixed-run body text.
+fn fixture(first: &str, style: Option<&str>) -> Document {
+    let mut d = Document::from_text(&format!("{first}\nBody paragraph."));
+    let p = d.para_mut(StoryRef::Body, &Path::top(0)).unwrap();
+    p.props.style = Some("Heading1".into());
+    p.props.outline_level = Some(0);
+    p.mark = CharProps { size: Some(40.0), bold: Some(true), ..Default::default() };
+    let p = d.para_mut(StoryRef::Body, &Path::top(1)).unwrap();
+    p.props = ParaProps {
+        style: style.map(str::to_string),
+        align: Some(Align::Right),
+        indent_left: Some(18.0),
+        space_after: Some(7.0),
+        keep_next: Some(false),
+        ..Default::default()
+    };
+    p.mark = CharProps { size: Some(11.0), bold: Some(false), lang: Some("en-GB".into()), ..Default::default() };
+    p.format(0, 4, &|c| c.italic = Some(true)).unwrap();
+    d
+}
+
+/// Compare paragraph, mark and inline-run formatting independently of text.
+fn assert_format(p: &Paragraph, expected: &Paragraph) {
+    assert_eq!(p.props, expected.props, "the following paragraph keeps all its properties");
+    assert_eq!(p.mark, expected.mark, "the empty heading's mark must not replace the following mark");
+    assert_eq!(serde_json::to_value(&p.runs).unwrap(), serde_json::to_value(&expected.runs).unwrap());
+}
+
+/// Deleting an empty paragraph preserves the following style, including implicit Normal.
+#[test]
+fn deleting_an_empty_paragraph_keeps_the_following_format() {
+    for first_style in ["Heading1", "Heading3", "Normal"] {
+        for style in [None, Some("Normal"), Some("Heading2")] {
+            let mut d = fixture("", style);
+            d.para_mut(StoryRef::Body, &Path::top(0)).unwrap().props.style = Some(first_style.into());
+            let expected = d.para_at(&Pos::body(1, 0)).unwrap().clone();
+            assert_eq!(d.delete_range(&Pos::body(0, 0), &Pos::body(1, 0)).unwrap(), Pos::body(0, 0));
+            let p = d.para_at(&Pos::body(0, 0)).unwrap();
+            assert_eq!(p.text, expected.text);
+            assert_format(p, &expected);
+        }
+    }
+}
+
+/// A selection spanning several paragraphs keeps the surviving tail format in both directions.
+#[test]
+fn an_empty_start_preserves_the_last_paragraph_over_a_multi_paragraph_selection() {
+    for backwards in [false, true] {
+        let mut d = fixture("", None);
+        d.body.insert(1, para_block(Paragraph::with_text("Removed middle paragraph", Default::default()).styled("Heading2")));
+        let mut expected = d.para_at(&Pos::body(2, 0)).unwrap().clone();
+        expected.delete(0, 5).unwrap();
+        let (a, b) = (Pos::body(0, 0), Pos::body(2, 5));
+        d.delete_range(if backwards { &b } else { &a }, if backwards { &a } else { &b }).unwrap();
+        assert_eq!(d.body.len(), 1);
+        let p = d.para_at(&Pos::body(0, 0)).unwrap();
+        assert_eq!(p.text, "paragraph.");
+        assert_format(p, &expected);
+    }
+}
+
+/// A control whose closing marker survives must move across the join without replacing tail formatting.
+#[test]
+fn an_empty_start_keeps_a_control_from_a_deleted_middle_paragraph() {
+    use wordcraft_doc::control::ContentControl;
+    use wordcraft_doc::{InlineObject, Revision, RevisionKind};
+
+    let mut d = fixture("", Some("Normal"));
+    d.body.insert(1, para_block(Paragraph::with_text("Removed", Default::default())));
+    d.body.push(para_block(Paragraph::with_text("After", Default::default())));
+    d.revisions.push(Revision { kind: RevisionKind::Insert, ..Default::default() });
+    d.revisions.push(Revision { kind: RevisionKind::Delete, ..Default::default() });
+    let p = d.para_mut(StoryRef::Body, &Path::top(2)).unwrap();
+    p.mark.ins = Some(0);
+    p.mark.del = Some(1);
+    let expected = p.clone();
+    let marker = CharProps { bold: Some(true), ins: Some(0), ..Default::default() };
+    d.insert_object(&Pos::body(1, 0), InlineObject::ControlStart { control: Box::new(ContentControl::default()) }, &marker).unwrap();
+    d.insert_object(&Pos::body(3, 5), InlineObject::ControlEnd, &Default::default()).unwrap();
+    assert_eq!(d.control_ranges(StoryRef::Body).len(), 1);
+
+    d.delete_range(&Pos::body(0, 0), &Pos::body(2, 0)).unwrap();
+    let p = d.para_at(&Pos::body(0, 0)).unwrap();
+    assert_eq!(p.props, expected.props);
+    assert_eq!(p.mark, expected.mark, "tail revision IDs and formatting survive the moved marker");
+    assert_eq!(p.props_of_char(0), &marker);
+    assert!(matches!(p.objects.as_slice(), [InlineObject::ControlStart { .. }]));
+    assert_eq!(p.text, format!("\u{fffc}{}", expected.text));
+    let ranges = d.control_ranges(StoryRef::Body);
+    assert_eq!(ranges.len(), 1);
+    assert_eq!(ranges[0].start, Pos::body(0, 0));
+    assert_eq!(ranges[0].end, Pos::body(1, 5));
+    assert!(!d.has_unbalanced_controls());
+}
+
+/// The empty-paragraph rule applies in header stories and table-cell containers.
+#[test]
+fn empty_paragraph_merges_work_in_a_header_and_a_table_cell() {
+    for cell in [false, true] {
+        let mut d = fixture("", Some("Normal"));
+        let expected = d.para_at(&Pos::body(1, 0)).unwrap().clone();
+        let (story, a, b) = if cell {
+            let mut table = Table::new(1, 1, 300.0);
+            table.rows[0].cells[0].blocks = d.body.clone();
+            d.body = vec![Arc::new(Block::Table(table)), para_block(Paragraph::new())];
+            (StoryRef::Body, Path(vec![0, 0, 0, 0]), Path(vec![0, 0, 0, 1]))
+        } else {
+            let id = d.add_part(wordcraft_doc::PartKind::Header, d.body.clone());
+            (StoryRef::Part(id), Path::top(0), Path::top(1))
+        };
+        d.delete_range(&Pos { story, path: a.clone(), off: 0 }, &Pos { story, path: b, off: 0 }).unwrap();
+        assert_format(d.para(story, &a).unwrap(), &expected);
+    }
+}
+
+/// Absorbing an empty section keeps the following format and inherited header.
+#[test]
+fn deleting_an_empty_section_paragraph_keeps_format_and_linked_headers() {
+    let mut d = fixture("", None);
+    let expected = d.para_at(&Pos::body(1, 0)).unwrap().clone();
+    let header = d.add_part(wordcraft_doc::PartKind::Header, vec![para_block(Paragraph::with_text("Header", Default::default()))]);
+    let mut section = d.last_section.clone();
+    section.headers.default = Some(header);
+    d.para_mut(StoryRef::Body, &Path::top(0)).unwrap().section = Some(Box::new(section));
+    d.last_section.margin_left = 96.0;
+    d.delete_range(&Pos::body(0, 0), &Pos::body(1, 0)).unwrap();
+    assert_format(d.para_at(&Pos::body(0, 0)).unwrap(), &expected);
+    assert!(d.para_at(&Pos::body(0, 0)).unwrap().section.is_none());
+    assert_eq!(d.last_section.headers.default, Some(header));
+    assert_eq!(d.last_section.margin_left, 96.0);
+}
+
+/// Deleting actual text continues to use the first paragraph merge rule.
+#[test]
+fn joining_nonempty_paragraphs_keeps_the_existing_first_paragraph_rule() {
+    for off in [0, 7] {
+        let mut d = fixture("Heading", None);
+        let expected = d.para_at(&Pos::body(0, 0)).unwrap().props.clone();
+        d.delete_range(&Pos::body(0, off), &Pos::body(1, 0)).unwrap();
+        assert_eq!(d.para_at(&Pos::body(0, 0)).unwrap().props, expected);
+    }
+}
+
+/// Removing a soft line break must not change paragraph formatting.
+#[test]
+fn a_soft_break_is_not_an_empty_paragraph_merge() {
+    let mut d = fixture("", None);
+    let p = d.para_mut(StoryRef::Body, &Path::top(0)).unwrap();
+    p.insert_text(0, "\nHeading", &Default::default()).unwrap();
+    let expected = p.props.clone();
+    d.delete_range(&Pos::body(0, 0), &Pos::body(0, 1)).unwrap();
+    assert_eq!(d.body.len(), 2);
+    assert_eq!(d.para_at(&Pos::body(0, 0)).unwrap().props, expected);
+}

crates/docbin/src/drawing.rs

@@ -211,7 +211,11 @@ fn leaf(sp: &Sp, w: f32, h: f32, float: Float, b: &Build, pics: &mut Pictures) -
         kind = ShapeKind::TextBox;
     }
     let line = kind == ShapeKind::Line;
-    let fill = if line || p.filled == Some(false) { None } else { Some(Fill::solid(p.fill_color.unwrap_or(Rgb::WHITE))) };
+    // Brackets and braces (85 to 88, bracket and brace pairs 185, 186) are unfilled unless
+    // `fFilled` says otherwise, as Word's VML shapetypes for them say (`filled="f"`). They are
+    // drawn as their box here, which a default fill would make an opaque block over the text.
+    let unfilled = p.filled == Some(false) || (matches!(sp.spt, 85..=88 | 185 | 186) && p.filled != Some(true));
+    let fill = if line || unfilled { None } else { Some(Fill::solid(p.fill_color.unwrap_or(Rgb::WHITE))) };
     let stroke = if p.line == Some(false) { None } else { Some(p.line_color.unwrap_or(Rgb::BLACK)) };
     // Line widths are EMUs; Word's default is 0.75 pt.
     let stroke_width = p.line_width.map_or(0.75, |e| (e as f32 / 12_700.0).clamp(0.0, 100.0));

crates/docbin/src/fkp.rs

@@ -7,9 +7,26 @@ use crate::DocbinError;
 /// Largest bin table (entries) we accept.
 const MAX_BINS: usize = 100_000;
 
+/// A Chpx grpprl and the FCs `[lo, hi)` for which it is the one [`Bins::chpx`] returns.
+#[derive(Clone, Copy)]
+pub(crate) struct ChpxRun<'w> {
+    pub(crate) grpprl: &'w [u8],
+    lo: u32,
+    hi: u32,
+}
+
+impl ChpxRun<'_> {
+    pub(crate) fn contains(&self, fc: u32) -> bool {
+        self.lo <= fc && fc < self.hi
+    }
+}
+
 /// A parsed `PlcBte*`: FC breakpoints and the FKP page each range lives in.
 pub(crate) struct Bins {
     fcs: Vec<u32>,
+    /// `fcs` ascends, so the bin of an FC holds every FC up to the next breakpoint (a corrupt
+    /// table that does not gets no cached ranges, [`ChpxRun`]).
+    sorted: bool,
     /// Page number (× 512 = byte offset in the WordDocument stream).
     pns: Vec<u32>,
 }
@@ -20,7 +37,7 @@ impl Bins {
         let fc = fc as usize;
         let lcb = lcb as usize;
         if lcb < 4 || !(lcb - 4).is_multiple_of(8) {
-            return Ok(Bins { fcs: Vec::new(), pns: Vec::new() });
+            return Ok(Bins { fcs: Vec::new(), sorted: true, pns: Vec::new() });
         }
         let end = fc.checked_add(lcb).ok_or_else(|| DocbinError::Malformed("bin table overflow".into()))?;
         let plc = table.get(fc..end).ok_or_else(|| DocbinError::Malformed("bin table outside the Table stream".into()))?;
@@ -38,36 +55,58 @@ impl Bins {
             // PnBte* keeps the page number in its low bits; the rest must be zero.
             pns.push(at(n + 1 + i).ok_or_else(|| DocbinError::Malformed("truncated bin aPn".into()))? & 0x003F_FFFF);
         }
-        Ok(Bins { fcs, pns })
+        let sorted = fcs.windows(2).all(|w| w.first() <= w.get(1));
+        Ok(Bins { fcs, sorted, pns })
     }
 
-    /// The page containing `fc`, as a slice of the WordDocument stream.
-    fn page<'w>(&self, word: &'w [u8], fc: u32) -> Option<&'w [u8]> {
+    /// The page containing `fc`, as a slice of the WordDocument stream, and the FC range
+    /// `[lo, hi)` of its bin (`hi` is `u32::MAX` for the last; an empty range when the table
+    /// is not ascending, so nothing is cached from it).
+    fn page<'w>(&self, word: &'w [u8], fc: u32) -> Option<(&'w [u8], u32, u32)> {
         let i = self.fcs.partition_point(|&f| f <= fc).checked_sub(1)?;
         let pn = *self.pns.get(i)?;
         let at = pn.checked_mul(512)? as usize;
-        word.get(at..at.checked_add(512)?)
+        let page = word.get(at..at.checked_add(512)?)?;
+        if !self.sorted {
+            return Some((page, 1, 0));
+        }
+        let lo = self.fcs.get(i).copied().unwrap_or(0);
+        let hi = self.fcs.get(i + 1).copied().unwrap_or(u32::MAX);
+        Some((page, lo, hi))
     }
 
     /// The grpprl of the Chpx covering `fc` ([MS-DOC] ChpxFkp: rgfc u32s, rgb offsets ×2,
     /// crun in the last byte).
     pub(crate) fn chpx<'w>(&self, word: &'w [u8], fc: u32) -> &'w [u8] {
-        let Some(page) = self.page(word, fc) else { return &[] };
-        let Some(&crun) = page.get(511) else { return &[] };
+        self.chpx_run(word, fc).grpprl
+    }
+
+    /// [`Bins::chpx`] with the FC range over which the same grpprl is the answer, so a walk
+    /// over consecutive characters looks it up once per run, not once per character.
+    pub(crate) fn chpx_run<'w>(&self, word: &'w [u8], fc: u32) -> ChpxRun<'w> {
+        let none = ChpxRun { grpprl: &[], lo: 1, hi: 0 };
+        let Some((page, page_lo, page_hi)) = self.page(word, fc) else { return none };
+        let Some(&crun) = page.get(511) else { return none };
         let crun = crun.min(0x65) as usize;
         let rgfc = |i: usize| -> Option<u32> { page.get(i * 4..i * 4 + 4).map(|b| u32::from_le_bytes([b[0], b[1], b[2], b[3]])) };
         // rgb array sits right after rgfc (crun+1 u32s); each byte is a half-offset into the
         // region between rgb's end and crun.
         let rgb_at = (crun + 1) * 4;
+        // The last run starting at or before `fc`. It stays the answer for every FC from its own
+        // start up to the smallest start after it (also in a page whose starts do not ascend).
         let j = (0..crun).rev().find(|&j| rgfc(j).is_some_and(|f| f <= fc));
-        let Some(j) = j else { return &[] };
-        let Some(&off) = page.get(rgb_at + j) else { return &[] };
+        let first_after = j.map_or(0, |j| j + 1);
+        let hi = (first_after..crun).filter_map(rgfc).min().map_or(page_hi, |m| m.min(page_hi));
+        let lo = j.and_then(rgfc).map_or(page_lo, |f| f.max(page_lo));
+        let run = |grpprl: &'w [u8]| ChpxRun { grpprl, lo, hi };
+        let Some(j) = j else { return run(&[]) };
+        let Some(&off) = page.get(rgb_at + j) else { return run(&[]) };
         if off == 0 {
-            return &[];
+            return run(&[]);
         }
         let at = off as usize * 2;
-        let Some(&cb) = page.get(at) else { return &[] };
-        page.get(at + 1..at + 1 + cb as usize).unwrap_or(&[])
+        let Some(&cb) = page.get(at) else { return run(&[]) };
+        run(page.get(at + 1..at + 1 + cb as usize).unwrap_or(&[]))
     }
 
     /// The (istd, grpprl) of the Papx of the paragraph whose mark is just before `fc`
@@ -75,7 +114,7 @@ impl Bins {
     /// entries (1 offset byte + 12 PHE bytes), crun in the last byte; the PAPX itself is a
     /// PapxInFkp with the cb/cb' size quirk). Query with the FC of the character after the mark.
     pub(crate) fn papx<'w>(&self, word: &'w [u8], fc: u32) -> (u16, &'w [u8]) {
-        let Some(page) = self.page(word, fc) else { return (0, &[]) };
+        let Some((page, _, _)) = self.page(word, fc) else { return (0, &[]) };
         let Some(&crun) = page.get(511) else { return (0, &[]) };
         // BX is 13 bytes; cap crun at what fits before byte 511.
         let max_run = (511usize.saturating_sub((crun as usize + 1) * 4)) / 13;

crates/docbin/src/lib.rs

@@ -260,6 +260,9 @@ fn read_limited(bytes: &[u8], password: Option<&str>, max_paragraphs: usize) ->
     };
 
     let mut doc = Document::new();
+    // Word opens Word 97-2003 files in compatibility mode 11 (Word 2003); the model's default
+    // 15 would apply Word 2013+ justification and line-breaking rules to them.
+    doc.settings.compat_mode = 11;
     if let Some(core) = metadata::read(&mut comp) {
         doc.core = core;
     }
@@ -502,29 +505,37 @@ struct ParaBuild {
     runs: Vec<Run>,
     props: wordcraft_doc::props::ParaProps,
     mark: CharProps,
-    /// The current run's properties, so runs only break when formatting changes.
-    cur: Option<CharProps>,
+    /// Which resolved formatting ([`FmtState::generation`]) the last character came from, while
+    /// the last run is still the one it extended: the same generation means equal properties.
+    run_gen: Option<u64>,
     /// Inline objects, each anchored at a U+FFFC pushed into `text`.
     objects: Vec<InlineObject>,
 }
 
 impl ParaBuild {
     fn push(&mut self, c: char, props: CharProps) {
+        self.run_gen = None;
         let len = c.len_utf8();
-        match &self.cur {
-            Some(p) if *p == props => {
-                if let Some(r) = self.runs.last_mut() {
-                    r.len += len;
-                }
-            }
-            _ => {
-                self.runs.push(Run { len, props: props.clone() });
-                self.cur = Some(props);
-            }
+        match self.runs.last_mut() {
+            Some(r) if r.props == props => r.len += len,
+            _ => self.runs.push(Run { len, props }),
         }
         self.text.push(c);
     }
 
+    /// [`ParaBuild::push`] for properties that are resolved once and shared by many characters:
+    /// they are cloned only when a run starts, and not even compared while `generation` stays
+    /// the same.
+    fn push_shared(&mut self, c: char, props: &CharProps, generation: u64) {
+        let len = c.len_utf8();
+        match self.runs.last_mut() {
+            Some(r) if self.run_gen == Some(generation) || r.props == *props => r.len += len,
+            _ => self.runs.push(Run { len, props: props.clone() }),
+        }
+        self.run_gen = Some(generation);
+        self.text.push(c);
+    }
+
     /// Append `c` to the current run, whose properties the caller knows it matches.
     fn extend(&mut self, c: char) {
         if let Some(r) = self.runs.last_mut() {
@@ -552,7 +563,7 @@ fn walk<'a>(ctx: &WalkCtx<'a>, cp_start: u32, cp_end: u32, pics: &mut media::Pic
     // Field assembly state: instruction characters are swallowed, the result is kept until
     // the end character, then emitted as one object (hyperlinks keep their result text).
     let mut field: Option<FieldBuild> = None;
-    let mut memo: Option<FmtMemo<'a>> = None;
+    let mut fmt = FmtState::default();
     while cp < cp_end {
         // Bookmark boundaries at this CP (main-document walks only; subdocument walks get
         // an empty table and `cp` never matches the main-document CPs).
@@ -715,16 +726,16 @@ fn walk<'a>(ctx: &WalkCtx<'a>, cp_start: u32, cp_end: u32, pics: &mut media::Pic
                     pb.objects.push(InlineObject::NoteRef { kind: *kind, id: *id, custom: String::new() });
                 }
             }
-            '\t' => push_formatted(&mut pb, cp, '\t', ctx, field.as_mut(), &mut memo),
-            '\u{B}' => push_formatted(&mut pb, cp, '\n', ctx, field.as_mut(), &mut memo),
-            '\u{C}' => push_formatted(&mut pb, cp, '\u{C}', ctx, field.as_mut(), &mut memo),
-            '\u{E}' => push_formatted(&mut pb, cp, '\u{E}', ctx, field.as_mut(), &mut memo),
-            '\u{1E}' => push_formatted(&mut pb, cp, '\u{2011}', ctx, field.as_mut(), &mut memo),
-            '\u{1F}' => push_formatted(&mut pb, cp, '\u{AD}', ctx, field.as_mut(), &mut memo),
+            '\t' => push_formatted(&mut pb, cp, '\t', ctx, field.as_mut(), &mut fmt),
+            '\u{B}' => push_formatted(&mut pb, cp, '\n', ctx, field.as_mut(), &mut fmt),
+            '\u{C}' => push_formatted(&mut pb, cp, '\u{C}', ctx, field.as_mut(), &mut fmt),
+            '\u{E}' => push_formatted(&mut pb, cp, '\u{E}', ctx, field.as_mut(), &mut fmt),
+            '\u{1E}' => push_formatted(&mut pb, cp, '\u{2011}', ctx, field.as_mut(), &mut fmt),
+            '\u{1F}' => push_formatted(&mut pb, cp, '\u{AD}', ctx, field.as_mut(), &mut fmt),
             // Object anchors, comment references and other special characters are dropped
             // (or handled by later passes).
             c if (c as u32) < 0x20 => {}
-            c => push_formatted(&mut pb, cp, c, ctx, field.as_mut(), &mut memo),
+            c => push_formatted(&mut pb, cp, c, ctx, field.as_mut(), &mut fmt),
         }
         cp = cp.saturating_add(units);
     }
@@ -854,34 +865,53 @@ fn num_of(ilfo: i32) -> Option<u32> {
 /// plus the piece's Prm, over the paragraph style's base character props. Inside a field,
 /// the character is recorded into the instruction or the cached result instead of (or as
 /// well as) the visible text.
-fn push_formatted<'a>(pb: &mut ParaBuild, cp: u32, c: char, ctx: &WalkCtx<'a>, field: Option<&mut FieldBuild>, memo: &mut Option<FmtMemo<'a>>) {
-    let fc = ctx.pieces.fc_of_cp(cp).unwrap_or(0);
-    let grpprl = ctx.chpx_bins.chpx(ctx.word, fc);
-    let prm = ctx.pieces.prm_of(cp);
-    let style = pb.props.style.as_deref();
-    let mut props = match memo.as_ref().filter(|m| std::ptr::eq(m.chpx, grpprl) && m.prm == prm && m.style.as_deref() == style) {
-        Some(m) => m.props.clone(),
+fn push_formatted<'a>(pb: &mut ParaBuild, cp: u32, c: char, ctx: &WalkCtx<'a>, field: Option<&mut FieldBuild>, st: &mut FmtState<'a>) {
+    // The piece and the CHPX run of the previous character usually cover this one too.
+    let span = match st.span.filter(|s| s.contains(cp)) {
+        Some(s) => Some(s),
         None => {
-            let default;
-            let base = match style.and_then(|id| ctx.sheet.get(id)) {
-                Some(st) => &st.chr,
-                None => {
-                    default = CharProps::default();
-                    &default
-                }
-            };
-            let mut props = char_props(grpprl, ctx.fonts, base);
-            apply_prm(&mut props, ctx.pieces.piece_prm(cp), ctx.fonts, base, ctx.work);
-            *memo = Some(FmtMemo { chpx: grpprl, prm, style: style.map(str::to_owned), props: props.clone() });
-            props
+            st.span = ctx.pieces.span(cp);
+            st.span
         }
     };
+    let fc = span.and_then(|s| s.fc_at(cp)).unwrap_or(0);
+    let run = match st.chpx.filter(|r| r.contains(fc)) {
+        Some(r) => r,
+        None => {
+            let r = ctx.chpx_bins.chpx_run(ctx.word, fc);
+            st.chpx = Some(r);
+            r
+        }
+    };
+    let grpprl = run.grpprl;
+    let prm = span.map_or(0, |s| s.prm());
+    let style = pb.props.style.as_deref();
+    let hit = st.memo.as_ref().is_some_and(|m| std::ptr::eq(m.chpx, grpprl) && m.prm == prm && m.style.as_deref() == style);
+    if !hit {
+        let default;
+        let base = match style.and_then(|id| ctx.sheet.get(id)) {
+            Some(sty) => &sty.chr,
+            None => {
+                default = CharProps::default();
+                &default
+            }
+        };
+        let mut props = char_props(grpprl, ctx.fonts, base);
+        apply_prm(&mut props, ctx.pieces.piece_prm(cp), ctx.fonts, base, ctx.work);
+        st.memo = Some(FmtMemo { chpx: grpprl, prm, style: style.map(str::to_owned), props });
+        st.generation += 1;
+    }
+    let Some(m) = st.memo.as_ref() else { return };
     // A symbol character (Insert › Symbol) is stored as the placeholder U+0028 with
     // sprmCSymbol carrying the font and the character code ([MS-DOC] §2.6.1).
+    let symbol = if c == '(' { symbol_operand(grpprl) } else { None };
+    if field.is_none() && symbol.is_none() {
+        pb.push_shared(c, &m.props, st.generation);
+        return;
+    }
+    let mut props = m.props.clone();
     let mut c = c;
-    if c == '('
-        && let Some((ftc, xchar)) = symbol_operand(grpprl)
-    {
+    if let Some((ftc, xchar)) = symbol {
         if let Some(font) = ctx.fonts.get(usize::from(ftc)) {
             props.font = Some(font.clone());
         }
@@ -900,6 +930,18 @@ fn push_formatted<'a>(pb: &mut ParaBuild, cp: u32, c: char, ctx: &WalkCtx<'a>, f
     }
 }
 
+/// What the walk carries from one character to the next while it resolves formatting: the piece
+/// and the CHPX run the last character was in (so most characters need no search), and the last
+/// formatting resolved.
+#[derive(Default)]
+struct FmtState<'a> {
+    span: Option<piece::PieceSpan<'a>>,
+    chpx: Option<fkp::ChpxRun<'a>>,
+    memo: Option<FmtMemo<'a>>,
+    /// Counts the memo's replacements, so equal values mean the same properties.
+    generation: u64,
+}
+
 /// The last character formatting resolved, reused while the CHPX, the piece's Prm and the
 /// paragraph style stay the same, so a Prm grpprl (up to 64 KB) is interpreted once per
 /// formatting change, not once per character (#3692).

crates/docbin/src/piece.rs

@@ -117,12 +117,6 @@ impl PieceTable {
         }
     }
 
-    /// The raw Prm of the piece holding `cp` (0, no properties, when there is none): two CPs
-    /// with the same value get the same [`PrmRef`].
-    pub(crate) fn prm_of(&self, cp: u32) -> u16 {
-        self.piece_of(cp).map_or(0, |p| p.prm)
-    }
-
     fn piece_of(&self, cp: u32) -> Option<&Pcd> {
         let i = self.cps.partition_point(|&c| c <= cp).checked_sub(1)?;
         if i >= self.pieces.len() {
@@ -143,6 +137,17 @@ impl PieceTable {
         if p.compressed { p.fc.checked_add(step)?.checked_div(2) } else { p.fc.checked_add(step) }
     }
 
+    /// The piece holding `cp`, with the CP range it covers: for every CP in that range the
+    /// answers of [`PieceTable::fc_of_cp`] and the piece's Prm come from this span alone,
+    /// so a sequential walk searches the piece table once per piece.
+    pub(crate) fn span(&self, cp: u32) -> Option<PieceSpan<'_>> {
+        let i = self.cps.partition_point(|&c| c <= cp).checked_sub(1)?;
+        let pcd = self.pieces.get(i)?;
+        let cp_lo = *self.cps.get(i)?;
+        let cp_hi = *self.cps.get(i + 1)?;
+        Some(PieceSpan { cp_lo, cp_hi, pcd })
+    }
+
     /// The CP just past the last piece: no text exists at or beyond it.
     pub(crate) fn cp_end(&self) -> u32 {
         self.cps.last().copied().unwrap_or(0)
@@ -187,6 +192,31 @@ impl PieceTable {
     }
 }
 
+/// One piece and the CPs `[cp_lo, cp_hi)` it covers ([`PieceTable::span`]).
+#[derive(Clone, Copy)]
+pub(crate) struct PieceSpan<'a> {
+    cp_lo: u32,
+    cp_hi: u32,
+    pcd: &'a Pcd,
+}
+
+impl PieceSpan<'_> {
+    pub(crate) fn contains(&self, cp: u32) -> bool {
+        self.cp_lo <= cp && cp < self.cp_hi
+    }
+
+    /// [`PieceTable::fc_of_cp`] for a CP inside the span.
+    pub(crate) fn fc_at(&self, cp: u32) -> Option<u32> {
+        let step = cp.checked_sub(self.cp_lo)?.checked_mul(2)?;
+        if self.pcd.compressed { self.pcd.fc.checked_add(step)?.checked_div(2) } else { self.pcd.fc.checked_add(step) }
+    }
+
+    /// The raw Prm of the piece: two CPs with the same value get the same [`PrmRef`].
+    pub(crate) fn prm(&self) -> u16 {
+        self.pcd.prm
+    }
+}
+
 /// What a piece's `Prm` refers to.
 pub(crate) enum PrmRef<'a> {
     None,

crates/docbin/src/tests.rs

@@ -337,6 +337,15 @@ fn empty_document_gets_a_paragraph() {
     assert_eq!(doc.body.len(), 1);
 }
 
+#[test]
+fn a_doc_opens_in_word_97_2003_compatibility_mode() {
+    let f = text_doc(&[Piece { compressed: false, text: "body\r".into() }], &[], 0);
+    let doc = read(&f).expect("opens");
+    // Word 97-2003 files are mode 11 in Word 2013+; mode 15 would change justification and
+    // line breaking.
+    assert_eq!(doc.settings.compat_mode, 11);
+}
+
 #[test]
 fn header_text_stays_out_of_body() {
     let f = text_doc(
@@ -2880,3 +2889,153 @@ fn field_results_keep_their_text_and_runs() {
         p.objects
     );
 }
+
+/// A one-page CHPX bin table over `page`: FCs 0..`end` live in page 1 of the returned stream.
+fn one_page_bins(page: &[u8], end: u32) -> (Vec<u8>, crate::fkp::Bins) {
+    let mut word = vec![0u8; 512];
+    word.extend_from_slice(page);
+    let plc = [0u32.to_le_bytes().as_slice(), end.to_le_bytes().as_slice(), 1u32.to_le_bytes().as_slice()].concat();
+    let bins = crate::fkp::Bins::parse(&plc, 0, 12).expect("bins");
+    (word, bins)
+}
+
+/// The CHPX run a lookup reports covers exactly the FCs that give the same grpprl: the walk
+/// reuses it for the following characters without asking again.
+#[test]
+fn a_chpx_run_covers_the_fcs_with_the_same_answer() {
+    let runs = [(0u32, grpprl(&[(0x0835, &[1])])), (5, Vec::new()), (9, grpprl(&[(0x0836, &[1])])), (30, grpprl(&[(0x0835, &[1])]))];
+    let (word, bins) = one_page_bins(&chpx_fkp(&runs, 60), 0x2000);
+    for fc in (fc_of(0) - 4)..(fc_of(70)) {
+        let run = bins.chpx_run(&word, fc);
+        assert_eq!(run.grpprl, bins.chpx(&word, fc), "fc {fc}");
+        for other in (fc_of(0) - 4)..(fc_of(70)) {
+            if run.contains(other) {
+                assert_eq!(bins.chpx(&word, other), run.grpprl, "fc {fc} run reused at {other}");
+            }
+        }
+    }
+    assert!(bins.chpx_run(&word, fc_of(6)).contains(fc_of(8)) && !bins.chpx_run(&word, fc_of(6)).contains(fc_of(9)));
+}
+
+/// A page whose run starts do not ascend (a corrupt file) gives the same answers as before and
+/// a range that never claims an FC with a different answer.
+#[test]
+fn a_chpx_run_is_sound_on_a_page_whose_starts_do_not_ascend() {
+    let runs = [(10u32, grpprl(&[(0x0835, &[1])])), (3, grpprl(&[(0x0836, &[1])])), (20, Vec::new()), (7, grpprl(&[(0x0835, &[1])]))];
+    let (word, bins) = one_page_bins(&chpx_fkp(&runs, 60), 0x2000);
+    for fc in (fc_of(0) - 4)..(fc_of(70)) {
+        let run = bins.chpx_run(&word, fc);
+        assert_eq!(run.grpprl, bins.chpx(&word, fc));
+        for other in (fc_of(0) - 4)..(fc_of(70)) {
+            if run.contains(other) {
+                assert_eq!(bins.chpx(&word, other), run.grpprl, "fc {fc} run reused at {other}");
+            }
+        }
+    }
+}
+
+/// A piece span answers exactly what the per-CP lookup answers, in every piece (compressed
+/// and UTF-16) and past the last.
+#[test]
+fn a_piece_span_matches_the_per_cp_lookups() {
+    let doc = text_doc(
+        &[
+            Piece { compressed: true, text: "abc".into() },
+            Piece { compressed: false, text: "d\u{e9}fgh".into() },
+            Piece { compressed: true, text: "ij".into() },
+        ],
+        &[],
+        0,
+    );
+    let mut comp = crate::open_container(&doc).expect("container");
+    let table = crate::stream(&mut comp, "0Table").expect("table");
+    let pieces = crate::piece::PieceTable::parse(&table, CLX_AT, table.len() as u32 - CLX_AT).expect("pieces");
+    for cp in 0..14 {
+        let span = pieces.span(cp);
+        assert_eq!(span.and_then(|s| s.fc_at(cp)), pieces.fc_of_cp(cp), "cp {cp}");
+        for other in 0..14 {
+            if span.is_some_and(|s| s.contains(other)) {
+                assert_eq!(pieces.span(other).and_then(|s| s.fc_at(other)), pieces.fc_of_cp(other));
+            }
+        }
+    }
+    assert!(pieces.span(10).is_none(), "past the last piece");
+}
+
+/// Characters of one formatting share one `CharProps`: it is cloned when a run starts, and not
+/// even compared while the formatting stays the same.
+#[test]
+fn shared_props_start_one_run_per_formatting() {
+    use wordcraft_doc::props::CharProps;
+    let bold = CharProps { bold: Some(true), ..Default::default() };
+    let mut pb = super::ParaBuild::default();
+    for _ in 0..100_000 {
+        pb.push_shared('a', &bold, 1);
+    }
+    pb.push_shared('b', &CharProps::default(), 2);
+    pb.push_shared('c', &bold, 3);
+    pb.push_shared('d', &bold, 4);
+    assert_eq!(pb.runs.iter().map(|r| r.len).collect::<Vec<_>>(), [100_000, 1, 2], "equal props under a new generation join the run");
+    pb.push('e', bold.clone());
+    pb.push_shared('f', &bold, 4);
+    assert_eq!(
+        pb.runs.iter().map(|r| r.len).collect::<Vec<_>>(),
+        [100_000, 1, 4],
+        "a plain push forgets the generation but still joins an equal run"
+    );
+    assert_eq!(pb.text.len(), 100_005);
+}
+
+/// Several bins, equal and descending breakpoints, a missing page and the last FC: wherever a
+/// run is reported as valid it gives the answer a fresh lookup gives.
+#[test]
+fn chpx_runs_are_sound_across_bins() {
+    let page_a = chpx_fkp(&[(0, grpprl(&[(0x0835, &[1])])), (8, grpprl(&[(0x0836, &[1])]))], 20);
+    let page_b = chpx_fkp(&[(20, Vec::new()), (26, grpprl(&[(0x0835, &[1])]))], 40);
+    let mut word = vec![0u8; 512];
+    word.extend_from_slice(&page_a);
+    word.extend_from_slice(&page_b);
+    let (a, b, end) = (fc_of(20), fc_of(40), fc_of(60));
+    let plc = |fcs: &[u32], pns: &[u32]| {
+        [fcs.iter().flat_map(|f| f.to_le_bytes()).collect::<Vec<u8>>(), pns.iter().flat_map(|p| p.to_le_bytes()).collect()].concat()
+    };
+    let tables = [
+        ("two bins", plc(&[0, a, end], &[1, 2])),
+        ("an empty bin", plc(&[0, a, a, end], &[1, 2, 2])),
+        ("a missing page", plc(&[0, a, end], &[1, 9])),
+        ("descending breakpoints", plc(&[0, b, a, end], &[1, 2, 1])),
+        ("fcs up to the maximum", plc(&[0, a, u32::MAX], &[1, 2])),
+    ];
+    for (what, t) in &tables {
+        let bins = crate::fkp::Bins::parse(t, 0, t.len() as u32).expect("bins");
+        let probes: Vec<u32> = (0..fc_of(50)).step_by(2).chain([u32::MAX - 1, u32::MAX]).collect();
+        for &fc in &probes {
+            let run = bins.chpx_run(&word, fc);
+            assert_eq!(run.grpprl, bins.chpx(&word, fc), "{what}: fc {fc}");
+            for &other in &probes {
+                if run.contains(other) {
+                    assert_eq!(bins.chpx(&word, other), run.grpprl, "{what}: fc {fc} run reused at {other}");
+                }
+            }
+        }
+    }
+}
+
+/// The same formatting changes through a whole read: runs break exactly where the CHPX changes
+/// (with a field and a symbol character in between), and nowhere else.
+#[test]
+fn formatting_runs_break_where_the_chpx_changes() {
+    let spec = RichSpec {
+        text: "aaaabbbb\u{13}PAGE\u{14}12\u{15}ccdd(ee\r",
+        styles: vec![(1, 0x0FFF, "Normal", Vec::new(), Vec::new())],
+        fonts: vec!["Times New Roman", "Arial"],
+        chpx_runs: vec![(0, grpprl(&[(0x0835, &[1])])), (4, Vec::new()), (8, grpprl(&[(0x0835, &[1])])), (20, grpprl(&[(0x0836, &[1])]))],
+        papx_entries: vec![(30, 0, Vec::new())],
+    };
+    let doc = read(&rich_doc(&spec)).expect("opens");
+    let p = para0(&doc);
+    let shape: Vec<(usize, Option<bool>, Option<bool>)> = p.runs.iter().map(|r| (r.len, r.props.bold, r.props.italic)).collect();
+    // "aaaa" bold, "bbbb" plain plus the field's anchor (3 bytes, default formatting), "ccd" bold
+    // (the run from the field to CP 20), "d(ee" italic.
+    assert_eq!(shape, [(4, Some(true), None), (7, None, None), (3, Some(true), None), (4, None, Some(true))], "{shape:?}");
+}

crates/docbin/src/tests/drawing.rs

@@ -183,6 +183,41 @@ fn shapes_keep_kind_colours_wrap_and_text_box_text() {
     assert_eq!(part.blocks.first().and_then(|b| b.as_para()).map(|p| p.text.as_str()), Some("Inside"));
 }
 
+#[test]
+fn braces_and_brackets_are_unfilled_unless_filled_is_set() {
+    // A right brace (88) in front of the text with no fill properties, and a left bracket (85)
+    // whose fFilled is set: only the bracket is filled. A plain rectangle (1) keeps its default
+    // white fill.
+    let brace = sp(&[fsp(88, 1026, 0xA00)]);
+    let bracket = sp(&[fsp(85, 1027, 0xA00), fopt(&[(0x01BF, (1 << 20) | (1 << 4))], &[])]);
+    let rect = sp(&[fsp(1, 1028, 0xA00)]);
+    let art = art_content(&[], &[brace, bracket, rect]);
+    let flags = spa_flags(0, 0, 3, false);
+    let spas =
+        plc(&[1, 2, 3, 5], &[spa(1026, [0, 0, 240, 1140], flags), spa(1027, [0, 0, 240, 1140], flags), spa(1028, [0, 0, 240, 1140], flags)].concat());
+    let text = "A\u{8}\u{8}\u{8}B\r";
+    let ccp = text.chars().count() as u32;
+    let spec = StructSpec { text, extra_pairs: vec![(40, spas), (50, art)], ..structured_spec(text, ccp) };
+    let doc = read(&struct_doc(&spec)).expect("opens");
+    let p0 = doc.body.first().and_then(|b| b.as_para()).expect("para");
+    let fills: Vec<_> = p0
+        .objects
+        .iter()
+        .map(|o| match o {
+            InlineObject::Shape { fill, stroke, w, h, .. } => (fill.clone(), *stroke, *w, *h),
+            other => panic!("{other:?}"),
+        })
+        .collect();
+    assert_eq!(
+        fills,
+        [
+            (None, Some(Rgb::BLACK), 12.0, 57.0),
+            (Some(Fill::solid(Rgb::WHITE)), Some(Rgb::BLACK), 12.0, 57.0),
+            (Some(Fill::solid(Rgb::WHITE)), Some(Rgb::BLACK), 12.0, 57.0),
+        ]
+    );
+}
+
 #[test]
 fn text_boxes_pointing_at_the_same_story_text_read_it_once() {
     // Two text boxes whose PlcftxbxTxt stories are both [0, 7) (the table steps back to CP 0

crates/docx/src/package.rs

@@ -350,6 +350,7 @@ impl Package {
                 let path = resolve(dir, target);
                 self.canonical(&path).map_or(path, str::to_string)
             };
+            rels.ids.entry(id.to_string()).or_insert(rels.list.len());
             rels.list.push(Rel { id: id.to_string(), kind: r.attr("Type").unwrap_or("").to_string(), target: resolved, external });
         }
         rels
@@ -405,11 +406,13 @@ pub struct Rel {
 #[derive(Clone, Debug, Default)]
 pub struct Rels {
     pub list: Vec<Rel>,
+    /// Each id's first relationship in `list`, so a lookup doesn't scan every relationship.
+    ids: HashMap<String, usize>,
 }
 
 impl Rels {
     pub fn by_id(&self, id: &str) -> Option<&Rel> {
-        self.list.iter().find(|r| r.id == id)
+        self.ids.get(id).and_then(|&i| self.list.get(i))
     }
     pub fn by_type(&self, full: &str) -> Option<&Rel> {
         self.list.iter().find(|r| rel_is(&r.kind, full))
@@ -547,6 +550,22 @@ mod tests {
         assert_eq!(crate::package_title(std::io::Cursor::new(z), 1 << 20).as_deref(), Some("Plan"));
     }
 
+    #[test]
+    fn relationships_are_looked_up_by_id_the_first_one_winning() {
+        let mut rels = String::from(r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships">"#);
+        for i in 0..20_000 {
+            rels.push_str(&format!(r#"<Relationship Id="rId{i}" Type="hyperlink" Target="https://example.com/{i}" TargetMode="External"/>"#));
+        }
+        rels.push_str(r#"<Relationship Id="rId7" Type="hyperlink" Target="https://example.com/dup" TargetMode="External"/></Relationships>"#);
+        let z = zip_of(&[("word/_rels/document.xml.rels", rels.into_bytes())]);
+        let r = Package::open(&z).unwrap().rels("word/document.xml");
+        assert_eq!(r.by_id("rId7").map(|r| r.target.as_str()), Some("https://example.com/7"), "the first rId7 wins");
+        assert_eq!(r.by_id("rId19999").map(|r| r.target.as_str()), Some("https://example.com/19999"));
+        assert!(r.by_id("rId20000").is_none());
+        // Hyperlink-style lookups, one per relationship, are each a map lookup.
+        assert_eq!((0..20_000).filter(|i| r.by_id(&format!("rId{i}")).is_some()).count(), 20_000);
+    }
+
     #[test]
     fn decompression_bomb_is_rejected() {
         let z = zip_of(&[("word/document.xml", vec![b' '; 2_000_000])]);

crates/docx/src/read/chart/model.rs

@@ -66,6 +66,9 @@ pub(super) struct Axis {
 
 pub(super) struct Series {
     pub(super) name: String,
+    /// The series' `c:idx` (an `xsd:unsignedInt`), which picks its automatic colour; `None`
+    /// when missing or not a valid number.
+    pub(super) idx: Option<usize>,
     pub(super) color: Option<Rgb>,
     pub(super) line_w: f32,
     /// Whether the series is drawn as a line (scatter series without one get markers).

crates/docx/src/read/chart/parse.rs

@@ -91,8 +91,11 @@ impl Chart {
         self.accents.get(i % 6).copied().unwrap_or(Rgb::BLACK)
     }
 
+    /// Colour of series `j`: its own, else the accent its `c:idx` picks (as Word does; `c:order`
+    /// doesn't matter), else the one its place in the file picks.
     pub(super) fn series_colour(&self, j: usize) -> Rgb {
-        self.series.get(j).and_then(|s| s.color).unwrap_or_else(|| self.accent(j))
+        let s = self.series.get(j);
+        s.and_then(|s| s.color).unwrap_or_else(|| self.accent(s.and_then(|s| s.idx).unwrap_or(j)))
     }
 
     /// Colour of point `i` of series `j`: its `c:dPt` override, else the varied accent (pie) or
@@ -199,6 +202,7 @@ pub(super) fn text_size(space: &El) -> f32 {
 impl Series {
     fn parse(ser: &El, scatter: bool, theme: &[Rgb], cap: usize) -> Series {
         let name = ser.child("c:tx").and_then(|t| t.find("c:v")).map(|v| label(v.text().trim())).unwrap_or_default();
+        let idx = ser.child("c:idx").and_then(|e| e.attr("val")).and_then(|v| v.trim().parse::<u32>().ok()).and_then(|v| usize::try_from(v).ok());
         let sp = ser.child("c:spPr");
         let ln = sp.and_then(|p| p.child("a:ln"));
         let marker = ser.child("c:marker").and_then(|m| m.child("c:spPr")).and_then(|p| fill_colour(p, theme));
@@ -219,7 +223,7 @@ impl Series {
             }
         }
         let labels = show_values(ser.child("c:dLbls"));
-        Series { name, color, line_w, line, values, xs, points, fmt, x_fmt, labels, raw: Vec::new() }
+        Series { name, idx, color, line_w, line, values, xs, points, fmt, x_fmt, labels, raw: Vec::new() }
     }
 }
 

crates/docx/src/read/chart/tests.rs

@@ -528,3 +528,65 @@ fn negative_sections_and_scientific_numbers() {
     assert_eq!(format_value(0.0, "0.00E+00", 1.0), "0.00E+00");
     assert_eq!(format_value(-12345.0, "0.0E-00", 1.0), "-1.2E04");
 }
+
+// An automatic series colour follows the series' `c:idx`, not its place in the file (#5963), as
+// measured in Word: `idx` 3 then 0 are accent4 then accent1, a lone `idx` 2 is accent3, and
+// `c:order` changes nothing.
+
+/// The fill colours of the bars of a column chart of `sers`, and of its legend swatches.
+fn column_colours(sers: &[String]) -> (Vec<Rgb>, Vec<Rgb>) {
+    let plot = format!(
+        r#"<c:barChart><c:barDir val="col"/><c:grouping val="clustered"/><c:varyColors val="0"/>{}<c:axId val="1"/><c:axId val="2"/></c:barChart><c:catAx><c:axId val="1"/></c:catAx><c:valAx><c:axId val="2"/></c:valAx>"#,
+        sers.concat()
+    );
+    let items = with_legend(&plot, "r", 300.0, 200.0);
+    let bars = filled(&items).iter().map(|b| b.0).collect();
+    let keys = items
+        .iter()
+        .filter_map(|it| if let GraphicItem::Shape { rect, fill, .. } = it { (rect[2] == SWATCH).then_some(*fill).flatten() } else { None })
+        .collect();
+    (bars, keys)
+}
+
+/// `ser` with its `c:idx` / `c:order` replaced by `idx_order` (an empty string drops them).
+fn renumbered(ser: &str, idx_order: &str) -> String {
+    let start = ser.find("<c:idx").expect("ser writes c:idx");
+    let end = ser.find("<c:tx>").expect("ser writes c:tx");
+    format!("{}{idx_order}{}", &ser[..start], &ser[end..])
+}
+
+#[test]
+fn automatic_series_colours_follow_c_idx() {
+    let accent = |k: usize| theme()[4 + k];
+    let a = ser(0, "A", None, &["Q1"], &["1"]);
+    let b = ser(1, "B", None, &["Q1"], &["2"]);
+    // idx 3 first, idx 0 second: accent4, accent1, in the bars and the legend.
+    let sers = [renumbered(&a, r#"<c:idx val="3"/><c:order val="0"/>"#), renumbered(&b, r#"<c:idx val="0"/><c:order val="1"/>"#)];
+    assert_eq!(column_colours(&sers), (vec![accent(3), accent(0)], vec![accent(3), accent(0)]));
+    // A lone series with idx 2 is accent3.
+    let lone = [renumbered(&a, r#"<c:idx val="2"/><c:order val="0"/>"#)];
+    assert_eq!(column_colours(&lone).0, [accent(2)]);
+    // c:order doesn't matter: idx 0 and 1 stay accent1 and accent2 with their orders swapped.
+    let swapped = [renumbered(&a, r#"<c:idx val="0"/><c:order val="1"/>"#), renumbered(&b, r#"<c:idx val="1"/><c:order val="0"/>"#)];
+    assert_eq!(column_colours(&swapped).0, [accent(0), accent(1)]);
+    // A series' own colour still wins.
+    let own = [renumbered(&ser(0, "A", Some("FF0000"), &["Q1"], &["1"]), r#"<c:idx val="3"/><c:order val="0"/>"#)];
+    assert_eq!(column_colours(&own).0, [Rgb(0xFF, 0, 0)]);
+}
+
+#[test]
+fn series_without_a_usable_c_idx_take_their_place() {
+    let accent = |k: usize| theme()[4 + k];
+    let a = ser(0, "A", None, &["Q1"], &["1"]);
+    let b = ser(1, "B", None, &["Q1"], &["2"]);
+    // No c:idx at all: the place in the file.
+    assert_eq!(column_colours(&[renumbered(&a, ""), renumbered(&b, "")]).0, [accent(0), accent(1)]);
+    // Junk, negative and out-of-range values: no panic, the place in the file.
+    for junk in ["x", "-1", "1e3", "99999999999999999999999"] {
+        let sers = [renumbered(&a, &format!(r#"<c:idx val="{junk}"/>"#)), renumbered(&b, &format!(r#"<c:idx val="{junk}"/>"#))];
+        assert_eq!(column_colours(&sers).0, [accent(0), accent(1)], "{junk:?}");
+    }
+    // The largest unsignedInt `c:idx` allows draws, without a panic.
+    let big = [renumbered(&a, &format!(r#"<c:idx val="{}"/>"#, u32::MAX))];
+    assert_eq!(column_colours(&big).0.len(), 1);
+}

crates/docx/src/read/embed.rs

@@ -239,7 +239,13 @@ impl Reader<'_> {
     /// Keep one part's bytes and content type (not its relationships).
     fn keep_part(&mut self, part: &str) -> bool {
         let lower = part.to_ascii_lowercase();
-        if !manifest_field_ok(part) || lower.ends_with(".rels") || lower == "[content_types].xml" || part.eq_ignore_ascii_case(&self.main) {
+        // `wordcraft:` names are WordCraft's own bookkeeping in `passthrough`, never a file's part.
+        if !manifest_field_ok(part)
+            || lower.ends_with(".rels")
+            || lower == "[content_types].xml"
+            || lower.starts_with("wordcraft:")
+            || part.eq_ignore_ascii_case(&self.main)
+        {
             log::warn!("docx: not carrying part {part:?} for an embedded object");
             return false;
         }
@@ -388,6 +394,23 @@ mod tests {
         assert!(roots.contains("keep.bin"), "the part is dropped: {roots}");
     }
 
+    /// A part named like WordCraft's own bookkeeping (`wordcraft:unmodelled.roots`) is not carried:
+    /// kept, it would be read back as that bookkeeping and add relationships the file chose.
+    #[test]
+    fn a_part_named_like_bookkeeping_is_not_carried() {
+        let forged = b"D\thttp://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument\tword/evil.xml\n";
+        let bytes = package(
+            r#"<w:p><w:r><w:t>text</w:t></w:r></w:p>"#,
+            &[("rIdDoc1", "/wordcraft:unmodelled.roots"), ("rIdDoc2", "/wordcraft:embedded.parts")],
+            &[("wordcraft:unmodelled.roots", forged), ("wordcraft:embedded.parts", b"P\tword/evil.xml\tapplication/xml\n")],
+        );
+        let doc = crate::read(&bytes).unwrap();
+        let roots = unmodelled_roots(&doc);
+        assert!(!roots.contains("wordcraft:") && !roots.contains("evil"), "{roots}");
+        let manifest = doc.passthrough.get(crate::package::EMBEDDED_PARTS).map(|v| String::from_utf8_lossy(v).to_string()).unwrap_or_default();
+        assert!(!manifest.contains("evil"), "{manifest}");
+    }
+
     /// An unqualified attribute is a relationship only where the markup names a relationship
     /// that way: a hyperlink's `id` names the part, a VML shape's `id` is its own.
     #[test]

crates/docx/src/read/parts.rs

@@ -55,6 +55,8 @@ impl Reader<'_> {
             .children("w:font")
             .filter(|f| f.child("w:family").and_then(|v| v.attr("w:val")) == Some("roman"))
             .filter_map(|f| f.attr("w:name"))
+            .flat_map(|n| n.split(';'))
+            .map(str::trim)
             .filter(|n| !n.is_empty() && n.len() <= MAX_FONT_NAME)
             .take(MAX_TABLE_FONTS)
             .map(str::to_string)

crates/docx/src/read/props.rs

@@ -5,16 +5,17 @@ use std::collections::HashMap;
 use wordcraft_doc::para::cap_alt;
 use wordcraft_doc::props::{
     Align, Border, BorderStyle, Borders, CellProps, CharProps, CombineBrackets, EastAsianLayout, HeightRule, Highlight, LineSpacing, NumRef,
-    ParaProps, Rgb, RowProps, TabAlign, TabLeader, TabStop, TableFloat, TableLook, TableProps, TextAlignment, TextColor, TextDirection, Underline,
-    VAlign, VMerge, VertAlign,
+    ParaProps, ParagraphTextDirection, Rgb, RowProps, TabAlign, TabLeader, TabStop, TableFloat, TableLook, TableProps, TextAlignment, TextColor,
+    TextDirection, Underline, VAlign, VMerge, VertAlign,
 };
 use wordcraft_doc::section::{
     Columns, DocGrid, DocGridType, LineNumberRestart, LineNumbering, MAX_CHAR_SPACE, MAX_GRID_PITCH, NotePos, NoteProps, NoteRestart, NumFormat,
     PageVAlign, SectionProps, SectionStart,
 };
 
+use super::MAX_FONT_NAME;
 use super::drawing_color::adjust_lum;
-use crate::units::{int, line_multiple, measure, on_off, tw, u32_of};
+use crate::units::{int, line_multiple, measure, on_off, tw, tw_i16, tw_u16, u32_of};
 
 /// A text frame (`w:framePr` without a drop cap, ECMA-376 §17.3.1.11), sizes and offsets in
 /// points (clamped), names kept as OOXML writes them.
@@ -145,6 +146,29 @@ pub(crate) fn resolve_level_only(sheet: &wordcraft_doc::styles::StyleSheet, p: &
     }
 }
 
+/// A `w:rFonts` value as the document keeps it: `None` when it names no family, else its leading
+/// families up to [`MAX_FONT_NAME`] bytes, cut at a `;` so no name is split. A LibreOffice file
+/// names its Latin font as a semicolon-separated fallback list, and a real list names a few
+/// families; a longer value is corrupt or hostile, and every piece of text drawn in it would pay
+/// for its whole length as the list is classified and resolved. Keeping the leading families keeps
+/// the family such a list picks, and the rest of the value is dropped like `w:cs`/`w:eastAsia`
+/// drop an over-long name.
+fn bounded_font_list(v: &str) -> Option<String> {
+    if v.len() <= MAX_FONT_NAME {
+        return (!v.is_empty()).then(|| v.to_string());
+    }
+    let mut end = 0usize;
+    for part in v.split_inclusive(';') {
+        let next = end + part.len();
+        if next > MAX_FONT_NAME {
+            break;
+        }
+        end = next;
+    }
+    let kept = v.get(..end)?.trim_end_matches(';');
+    (!kept.is_empty()).then(|| kept.to_string())
+}
+
 impl PropCtx {
     pub fn style_id(&self, id: &str) -> String {
         self.alias.get(id).cloned().unwrap_or_else(|| id.to_string())
@@ -226,6 +250,8 @@ impl PropCtx {
         shd_fill(e)
     }
 
+    /// The character properties `e` (a `w:rPr`) sets: direct formatting, plus the fonts, size,
+    /// weight and posture of the complex-script and East Asian text beside them.
     pub fn rpr(&self, e: &El) -> CharProps {
         let mut c = CharProps::default();
         for k in e.els() {
@@ -234,7 +260,11 @@ impl PropCtx {
                 "w:rFonts" => {
                     // A theme font supersedes the fixed font beside it (`w:asciiTheme` over
                     // `w:ascii`, `w:hAnsiTheme` over `w:hAnsi`).
-                    let named = |a: &str| k.attr(a).filter(|f| !f.is_empty()).map(str::to_string);
+                    // A LibreOffice file names the Latin font as a semicolon-separated fallback
+                    // list. Names are bounded like `w:cs`/`w:eastAsia` below: a longer one is
+                    // corrupt or hostile, and every piece of text drawn in it would pay for its
+                    // whole length as the list is classified (see `bounded_font_list`).
+                    let named = |a: &str| k.attr(a).and_then(bounded_font_list);
                     let themed = |a: &str| k.attr(a).and_then(|t| self.theme_font(t));
                     // `w:eastAsia` and `w:cs` name fonts for other scripts and stay in `font_ea` /
                     // `font_cs`: `w:hint` only picks between them for ambiguous characters, never for
@@ -338,6 +368,7 @@ impl PropCtx {
                 "w:suppressAutoHyphens" => p.suppress_hyphens = Some(on_off(k)),
                 "w:suppressLineNumbers" => p.suppress_line_numbers = Some(on_off(k)),
                 "w:bidi" => p.bidi = Some(on_off(k)),
+                "w:textDirection" => p.text_direction = k.attr("w:val").and_then(ParagraphTextDirection::from_ooxml),
                 "w:kinsoku" => p.kinsoku = Some(on_off(k)),
                 "w:wordWrap" => p.word_wrap = Some(on_off(k)),
                 "w:overflowPunct" => p.overflow_punct = Some(on_off(k)),
@@ -376,10 +407,10 @@ impl PropCtx {
                     p.tabs = Some(tabs);
                 }
                 "w:spacing" => {
-                    if let Some(v) = tw(k, "w:before") {
+                    if let Some(v) = tw_u16(k, "w:before") {
                         p.space_before = Some(v.max(0.0));
                     }
-                    if let Some(v) = tw(k, "w:after") {
+                    if let Some(v) = tw_u16(k, "w:after") {
                         p.space_after = Some(v.max(0.0));
                     }
                     // HTML-style automatic spacing; "0" switches an inherited one off.
@@ -402,15 +433,15 @@ impl PropCtx {
                     }
                 }
                 "w:ind" => {
-                    if let Some(v) = tw(k, "w:left").or_else(|| tw(k, "w:start")) {
+                    if let Some(v) = tw_i16(k, "w:left").or_else(|| tw_i16(k, "w:start")) {
                         p.indent_left = Some(v);
                     }
-                    if let Some(v) = tw(k, "w:right").or_else(|| tw(k, "w:end")) {
+                    if let Some(v) = tw_i16(k, "w:right").or_else(|| tw_i16(k, "w:end")) {
                         p.indent_right = Some(v);
                     }
-                    if let Some(v) = tw(k, "w:hanging") {
+                    if let Some(v) = tw_i16(k, "w:hanging") {
                         p.indent_first = Some(-v);
-                    } else if let Some(v) = tw(k, "w:firstLine") {
+                    } else if let Some(v) = tw_i16(k, "w:firstLine") {
                         p.indent_first = Some(v);
                     }
                     // Character units (hundredths of a character), which win when non-zero.

crates/docx/src/read/story/drawing.rs

@@ -41,6 +41,7 @@ impl Reader<'_> {
             }
             let source = self.keep_embedded(c.child("a:graphic")?, rels, &[])?;
             float.set_spin(xfrm_spin(content.child("w14:xfrm")));
+            float.bare_turn = !anchored && float.rot != 0.0 && float.effect == [0.0; 4];
             let (px, py) = float.spin_pad(w, h);
             for (e, p) in float.effect.iter_mut().zip([px, py, px, py]) {
                 if *e != 0.0 {
@@ -106,7 +107,11 @@ impl Reader<'_> {
         self.set_shape_id(&mut obj, c.child("wp:docPr").and_then(|p| p.attr("id")));
         // Word's effect extent also covers a rotated object's overhang; the model keeps that
         // apart (it follows from the angle), so take it back out.
+        // Shadow or glow room the file didn't record: saving records it (and the turn's room).
+        let fx_room = matches!(&obj, InlineObject::Shape { effects, .. } if effects.extent() != [0.0; 4]);
         if let Some(f) = obj.float_mut() {
+            // None recorded: Word keeps an inline one's line to the frame (see `Float::bare_turn`).
+            f.bare_turn = !anchored && !fx_room && f.rot != 0.0 && f.effect == [0.0; 4];
             let (px, py) = f.spin_pad(w, h);
             for (e, p) in f.effect.iter_mut().zip([px, py, px, py]) {
                 // Zero adds no effects room when rotation contracts this axis.
@@ -118,9 +123,33 @@ impl Reader<'_> {
         Some(obj)
     }
 
+    /// The hyperlink Word keeps on a drawing itself: its `wp:docPr`'s `a:hlinkClick` (ECMA-376
+    /// Part 1 §20.4.2.5, §21.1.2.3.5) naming an external hyperlink relationship. Word stores a
+    /// picture's link there rather than in a `w:hyperlink` around its run (#5267). Its `#name`
+    /// part, when the address has one, stays part of the address.
+    pub(super) fn drawing_link(d: &El, rels: &Rels) -> Option<String> {
+        let frame = d.child("wp:inline").or_else(|| d.child("wp:anchor"))?;
+        let id = frame.child("wp:docPr")?.child("a:hlinkClick")?.attr("r:id")?;
+        let rel = rels.by_id(id).filter(|r| r.external && crate::package::rel_is(&r.kind, rt::HYPERLINK))?;
+        Some(rel.target.trim().to_string()).filter(|u| !u.is_empty())
+    }
+
+    /// Native ink or an Office 2016 chart kept with its fallback (`mc:AlternateContent`), and the
+    /// [`Self::drawing_link`] of the choice it was read from, else of its fallback: another
+    /// choice's link is not this object's.
+    pub(super) fn read_alternate(&mut self, sc: &mut StoryCtx, ac: &El, rels: &Rels) -> Option<(InlineObject, Option<String>)> {
+        let (object, choice) = match self.read_ink_alternate(sc, ac, rels) {
+            Some(o) => (o, children_of_choice(ac).filter(|e| e.name == "mc:Choice")),
+            None => (self.read_chartex_alternate(sc, ac, rels)?, chartex_choice(ac)),
+        };
+        let link =
+            [choice, ac.child("mc:Fallback")].into_iter().flatten().filter_map(|e| e.child("w:drawing")).find_map(|d| Self::drawing_link(d, rels));
+        Some((object, link))
+    }
+
     /// Keep the selected native-ink branch and its picture fallback as one object. Reading only
     /// the fallback flattens the ink; reading only the choice loses its preview on the next save.
-    pub(super) fn read_ink_alternate(&mut self, sc: &mut StoryCtx, ac: &El, rels: &Rels) -> Option<InlineObject> {
+    fn read_ink_alternate(&mut self, sc: &mut StoryCtx, ac: &El, rels: &Rels) -> Option<InlineObject> {
         let choice = children_of_choice(ac).filter(|e| e.name == "mc:Choice")?;
         let drawing = choice.child("w:drawing")?;
         if drawing.find("a:graphicData")?.attr("uri") != Some(INK_URI) {
@@ -149,10 +178,8 @@ impl Reader<'_> {
     /// An Office 2016 chart (`cx:chart`: waterfall, sunburst, box & whisker…) drawn from its
     /// fallback picture, keeping both branches so saving writes the chart back rather than only
     /// the picture. `None` (read the fallback alone) when it can't be kept whole.
-    pub(super) fn read_chartex_alternate(&mut self, sc: &mut StoryCtx, ac: &El, rels: &Rels) -> Option<InlineObject> {
-        let choice = ac
-            .children("mc:Choice")
-            .find(|c| c.child("w:drawing").and_then(|d| d.find("a:graphicData")).and_then(|g| g.attr("uri")) == Some(CHARTEX_URI))?;
+    fn read_chartex_alternate(&mut self, sc: &mut StoryCtx, ac: &El, rels: &Rels) -> Option<InlineObject> {
+        let choice = chartex_choice(ac)?;
         // A requirement in a namespace we can't name couldn't be written back.
         let requires = choice.attr("Requires").unwrap_or("");
         if requires.split_whitespace().next().is_none() || requires.split_whitespace().any(|p| p == "?") {
@@ -563,6 +590,11 @@ impl Reader<'_> {
     }
 }
 
+/// The `mc:Choice` holding an Office 2016 chart (`cx:chart`) in an `mc:AlternateContent`.
+fn chartex_choice(ac: &El) -> Option<&El> {
+    ac.children("mc:Choice").find(|c| c.child("w:drawing").and_then(|d| d.find("a:graphicData")).and_then(|g| g.attr("uri")) == Some(CHARTEX_URI))
+}
+
 /// The alternative text of a `wp:docPr`, `pic:cNvPr` or `wps:cNvPr` (ECMA-376
 /// `CT_NonVisualDrawingProps`): the description (the title when there is none, as before), and
 /// the title and decorative mark (the `adec:decorative` extension of [MS-ODRAWXML]).

crates/docx/src/read/story/runs.rs

@@ -98,9 +98,9 @@ impl Reader<'_> {
                 }
             }
             "mc:AlternateContent" => {
-                if let Some(o) = self.read_ink_alternate(sc, k, rels).or_else(|| self.read_chartex_alternate(sc, k, rels)) {
+                if let Some((o, link)) = self.read_alternate(sc, k, rels) {
                     let props = self.run_props_none(ctx);
-                    self.emit_obj(sc, pb, o, &props);
+                    self.emit_linked_obj(sc, pb, o, &props, link);
                 } else if let Some(c) = children_of_choice(k) {
                     self.read_inline_children(sc, pb, c, rels, ctx, depth + 1);
                 }
@@ -237,7 +237,7 @@ impl Reader<'_> {
             },
             "w:drawing" => {
                 if let Some(o) = self.read_drawing(sc, k, rels) {
-                    self.emit_obj(sc, pb, o, props);
+                    self.emit_linked_obj(sc, pb, o, props, Self::drawing_link(k, rels));
                 }
             }
             "w:pict" => {
@@ -298,8 +298,8 @@ impl Reader<'_> {
             ),
             "w:ruby" => self.read_ruby(sc, pb, k, props),
             "mc:AlternateContent" => {
-                if let Some(o) = self.read_ink_alternate(sc, k, rels).or_else(|| self.read_chartex_alternate(sc, k, rels)) {
-                    self.emit_obj(sc, pb, o, props);
+                if let Some((o, link)) = self.read_alternate(sc, k, rels) {
+                    self.emit_linked_obj(sc, pb, o, props, link);
                 } else if let Some(c) = children_of_choice(k) {
                     for x in c.els() {
                         self.read_run_child(sc, pb, x, (props, rels), note, depth + 1);
@@ -376,6 +376,14 @@ impl Reader<'_> {
         }
     }
 
+    /// [`Self::emit_obj`] for a drawing that carries its own `link`: a link around the run wins.
+    fn emit_linked_obj(&mut self, sc: &mut StoryCtx, pb: &mut PB, o: InlineObject, props: &CharProps, link: Option<String>) {
+        match link.filter(|_| props.link.is_none()) {
+            Some(link) => self.emit_obj(sc, pb, o, &CharProps { link: Some(link), ..props.clone() }),
+            None => self.emit_obj(sc, pb, o, props),
+        }
+    }
+
     pub(super) fn emit_obj(&mut self, sc: &mut StoryCtx, pb: &mut PB, o: InlineObject, props: &CharProps) {
         match sink(sc) {
             Some(f) => {

crates/docx/src/read/story/vml.rs

@@ -224,6 +224,15 @@ impl Reader<'_> {
         {
             let id = self.read_textbox(sc, t, rels);
             if let Some(part) = self.doc.parts.get_mut(&id) {
+                // Text direction (#5193), as Word reads it: `mso-layout-flow-alt:bottom-to-top` is
+                // `vert270` whatever `layout-flow` says; otherwise `layout-flow:vertical…` (also
+                // `vertical-ideographic`, our `eaVert`) is `vert`, and anything else horizontal.
+                let tb_style = tb.attr("style").unwrap_or("");
+                if vml_style(tb_style, "mso-layout-flow-alt").is_some_and(|v| v.eq_ignore_ascii_case("bottom-to-top")) {
+                    part.body.vert = wordcraft_doc::props::TextVert::Vert270;
+                } else if vml_style(tb_style, "layout-flow").is_some_and(|v| v.to_ascii_lowercase().starts_with("vertical")) {
+                    part.body.vert = wordcraft_doc::props::TextVert::Vert;
+                }
                 // Internal margins `inset="l,t,r,b"` (ECMA-376 Part 4 §19.1.2.22), each one left
                 // out keeping Word's default; a bare number is in EMUs, VML's base unit.
                 // `o:insetmode="auto"` leaves them to the application.

crates/docx/src/read/styles.rs

@@ -2,6 +2,7 @@ use super::{
     AbstractNum, DocxError, El, Level, LevelSuffix, Num, NumFormat, PartFormat, Reader, Style, StyleKind, StyleSheet, TablePart, TableStyleParts,
     VAlign, flag, int, props, tw, u32_of,
 };
+use crate::units::tw_i16;
 use std::sync::Arc;
 
 /// Built-in style names Word stores in lower case, and how the UI shows them.
@@ -317,12 +318,12 @@ impl Reader<'_> {
             lv.align = props::align(j);
         }
         if let Some(ind) = l.child("w:pPr").and_then(|p| p.child("w:ind")) {
-            if let Some(v) = tw(ind, "w:left").or_else(|| tw(ind, "w:start")) {
+            if let Some(v) = tw_i16(ind, "w:left").or_else(|| tw_i16(ind, "w:start")) {
                 lv.indent = v;
             }
-            if let Some(h) = tw(ind, "w:hanging") {
+            if let Some(h) = tw_i16(ind, "w:hanging") {
                 lv.hanging = h;
-            } else if let Some(f) = tw(ind, "w:firstLine") {
+            } else if let Some(f) = tw_i16(ind, "w:firstLine") {
                 lv.hanging = -f;
             }
         }

crates/docx/src/read/toc.rs

@@ -83,13 +83,14 @@ pub(super) fn toc_control_to_toc(body: &mut Blocks) {
     }
     // An empty paragraph opening the control takes the field. It may end a section, as above
     // entries in columns (#3889).
-    if engine.get(s).and_then(|b| b.as_para()).is_some_and(|p| p.is_empty()) {
+    let empty_heading = engine.get(s).and_then(|b| b.as_para()).is_some_and(|p| p.is_empty());
+    if empty_heading {
         field_to_heading(&mut engine, s + 1);
     } else {
         toc_field_to_heading(&mut engine);
     }
     if crate::write::written_toc_span(&engine) == Some((s, last)) {
-        *body = toc;
+        *body = if empty_heading { toc } else { engine };
     }
 }
 
@@ -97,7 +98,7 @@ pub(super) fn toc_control_to_toc(body: &mut Blocks) {
 /// reproduce exactly. Other imported TOCs keep their explicit boundaries, independent of styles.
 pub(super) fn toc_field_to_heading(body: &mut Blocks) {
     for i in 1..body.len() {
-        if body[i - 1].as_para().is_some_and(|p| p.props.style.as_deref() == Some("TOCHeading")) {
+        if body.get(i - 1).and_then(|b| b.as_para()).is_some_and(|p| p.props.style.as_deref() == Some("TOCHeading")) {
             field_to_heading(body, i);
         }
     }

crates/docx/src/units.rs

@@ -116,6 +116,27 @@ pub fn tw(e: &El, attr: &str) -> Option<f32> {
     e.attr(attr).and_then(|v| measure(v, 20.0))
 }
 
+/// Signed twips Word keeps in 16 bits (paragraph and list indents, ECMA-376 Part 1 §17.3.1.12):
+/// a plain integer outside `i16` keeps its low 16 bits, two's complement, as Word reads it
+/// (66256 → 720). Values with a unit or a fraction read as [`tw`].
+pub fn tw_i16(e: &El, attr: &str) -> Option<f32> {
+    let v = e.attr(attr)?;
+    match v.trim().parse::<i64>() {
+        Ok(i) => Some(f32::from(i as i16) / 20.0),
+        Err(_) => measure(v, 20.0),
+    }
+}
+
+/// Unsigned twips Word keeps in 16 bits (paragraph spacing, §17.3.1.33): a plain integer above
+/// 65535 keeps its low 16 bits (65537 → 1). Other values read as [`tw`].
+pub fn tw_u16(e: &El, attr: &str) -> Option<f32> {
+    let v = e.attr(attr)?;
+    match v.trim().parse::<i64>() {
+        Ok(i) if i > 0xFFFF => Some(f32::from((i & 0xFFFF) as u16) / 20.0),
+        _ => measure(v, 20.0),
+    }
+}
+
 /// ST_OnOff: absent `w:val` = on; `0`, `false`, `off` = off.
 pub fn on_off(e: &El) -> bool {
     match e.attr("w:val") {
@@ -148,6 +169,24 @@ pub fn emu(pt: f32) -> String {
 mod tests {
     use super::*;
 
+    #[test]
+    fn twips_beyond_16_bits_wrap() {
+        let el = |a: &str| El { attrs: vec![("w:v".into(), a.into())], ..El::default() };
+        let s = |a: &str| tw_i16(&el(a), "w:v");
+        let u = |a: &str| tw_u16(&el(a), "w:v");
+        assert_eq!(s("66256"), Some(36.0));
+        assert_eq!(s("-65516"), Some(1.0));
+        assert_eq!(s("98304"), Some(-1638.4));
+        assert_eq!(s("32767"), Some(1638.35));
+        assert_eq!(s("-32768"), Some(-1638.4));
+        assert_eq!(s("1in"), Some(72.0));
+        assert_eq!(s("x"), None);
+        assert_eq!(u("65537"), Some(0.05));
+        assert_eq!(u("65535"), Some(3276.75));
+        assert_eq!(u("12pt"), Some(12.0));
+        assert_eq!(u("-20"), Some(-1.0));
+    }
+
     #[test]
     fn measures() {
         assert_eq!(measure("1440", 20.0), Some(72.0));

crates/docx/src/write/embed.rs

@@ -499,8 +499,9 @@ impl super::Writer<'_> {
                 }
             }
             "wp:effectExtent" => {
-                // Word's effect extent also covers a rotated object's overhang (its rotated bounds).
-                let (px, py) = float.spin_pad(w, h);
+                // Word's effect extent also covers a rotated object's overhang (its rotated bounds),
+                // unless the file it came from kept none.
+                let (px, py) = if float.is_bare_turn() { (0.0, 0.0) } else { float.spin_pad(w, h) };
                 let [el, et, er, eb] = float.effect_extent();
                 let ext = [el + px, et + py, er + px, eb + py].map(emu);
                 e.attrs.retain(|(k, _)| !matches!(k.as_str(), "l" | "t" | "r" | "b"));

crates/docx/src/write/props.rs

@@ -287,6 +287,9 @@ pub fn ppr_inner(w: &mut W, p: &ParaProps, framed: bool, num_change: Option<&Cha
     if let Some(a) = p.align {
         w.val("w:jc", align_val(a));
     }
+    if let Some(direction) = p.text_direction {
+        w.val("w:textDirection", direction.ooxml());
+    }
     if let Some(t) = p.text_alignment {
         w.val("w:textAlignment", t.ooxml());
     }

crates/docx/src/write/sdt.rs

@@ -86,11 +86,28 @@ pub fn plan(bl: &Blocks) -> Plan<'_> {
                 && (0..off).step_by(OBJ.len_utf8()).all(|o| matches!(p.object_at(o), Some(wordcraft_doc::InlineObject::ControlStart { .. })))
         })
     };
+    // Bookmark, comment and permission markers after the end may stay there: a body-level
+    // bookmark after a block control (Word's `_GoBack`) is read onto the end of its paragraph,
+    // and the control still covers the whole paragraph.
     let at_end = |b: usize, off: usize| {
+        use wordcraft_doc::InlineObject as O;
         para(b).is_some_and(|p| {
             let from = off + OBJ.len_utf8();
             p.text.get(from..).is_some_and(|t| t.chars().all(|c| c == OBJ))
-                && (from..p.len()).step_by(OBJ.len_utf8()).all(|o| matches!(p.object_at(o), Some(wordcraft_doc::InlineObject::ControlEnd)))
+                && (from..p.len()).step_by(OBJ.len_utf8()).all(|o| {
+                    matches!(
+                        p.object_at(o),
+                        Some(
+                            O::ControlEnd
+                                | O::BookmarkStart { .. }
+                                | O::BookmarkEnd { .. }
+                                | O::CommentStart { .. }
+                                | O::CommentEnd { .. }
+                                | O::PermStart { .. }
+                                | O::PermEnd { .. }
+                        )
+                    )
+                })
         })
     };
     let mut mode = vec![Mode::Inline; ranges.len()];

crates/docx/src/write/story/drawing.rs

@@ -10,9 +10,19 @@ pub(super) struct DrawingInfo<'a> {
     pub docpr: &'a str,
     pub name: &'a str,
     pub alt: &'a str,
+    /// The relationship id of the drawing's hyperlink (`a:hlinkClick`), see [`Writer::drawing_link`].
+    pub link: Option<&'a str>,
 }
 
 impl Writer<'_> {
+    /// The relationship of the hyperlink a drawing in a run with `props` carries on itself, as
+    /// Word writes a picture's link (`wp:docPr/a:hlinkClick`, #5267): a web or file address. A
+    /// link to a place in the document (`#name`) is only the `w:hyperlink` around the run.
+    pub(super) fn drawing_link(props: &wordcraft_doc::CharProps, rels: &mut PartRels) -> Option<String> {
+        let url = props.link.as_deref().filter(|l| !l.is_empty() && !l.starts_with('#'))?;
+        Some(rels.add(rt::HYPERLINK, url, true))
+    }
+
     /// A picture's `pic:pic` (the image in `media/{file}`), at `off` in its group (or 0, 0).
     pub(super) fn pic(&mut self, w: &mut W, o: &InlineObject, file: &str, off: (f32, f32), rels: &mut PartRels) {
         let InlineObject::Image { media, w: iw, h: ih, alt, crop, float, fx, .. } = o else { return };
@@ -20,7 +30,7 @@ impl Writer<'_> {
         self.used_media.insert(media.clone());
         w.open("pic:pic", &[]);
         w.open("pic:nvPicPr", &[]);
-        nv_props(w, "pic:cNvPr", "0", file, alt, &float.alt, float.hidden);
+        nv_props(w, "pic:cNvPr", ("0", file), alt, &float.alt, float.hidden, None);
         w.open("pic:cNvPicPr", &[]);
         w.empty("a:picLocks", &[("noChangeAspect", "1"), ("noChangeArrowheads", "1")]);
         w.close("pic:cNvPicPr");
@@ -66,7 +76,7 @@ impl Writer<'_> {
         w.open("wps:wsp", &[]);
         if let Some(id) = id {
             let name = shape_name(*kind, freeform.as_deref(), id);
-            nv_props(w, "wps:cNvPr", id, &name, &float.alt.text, &float.alt, float.hidden);
+            nv_props(w, "wps:cNvPr", (id, &name), &float.alt.text, &float.alt, float.hidden, None);
         }
         if kind.is_connector() {
             // A connector's glued ends (ECMA-376 §20.1.2.2.8 `cNvCxnSpPr`).
@@ -235,7 +245,7 @@ impl Writer<'_> {
     }
 
     pub(super) fn drawing_open(&mut self, w: &mut W, info: DrawingInfo<'_>) {
-        let DrawingInfo { float, size: (iw, ih), docpr, name, alt } = info;
+        let DrawingInfo { float, size: (iw, ih), docpr, name, alt, link } = info;
         let (cx, cy) = (emu(iw.max(0.0)), emu(ih.max(0.0)));
         if float.wrap == Wrap::Inline {
             w.open("wp:inline", &[("distT", "0"), ("distB", "0"), ("distL", "0"), ("distR", "0")]);
@@ -288,8 +298,9 @@ impl Writer<'_> {
             }
         }
         w.empty("wp:extent", &[("cx", &cx), ("cy", &cy)]);
-        // Word's effect extent also covers a rotated object's overhang (its rotated bounds).
-        let (px, py) = float.spin_pad(iw, ih);
+        // Word's effect extent also covers a rotated object's overhang (its rotated bounds), unless
+        // the file it came from kept none.
+        let (px, py) = if float.is_bare_turn() { (0.0, 0.0) } else { float.spin_pad(iw, ih) };
         let [el, et, er, eb] = float.effect_extent();
         let [el, et, er, eb] = [el + px, et + py, er + px, eb + py].map(emu);
         w.empty("wp:effectExtent", &[("l", &el), ("t", &et), ("r", &er), ("b", &eb)]);
@@ -316,7 +327,7 @@ impl Writer<'_> {
             Wrap::TopAndBottom => w.empty("wp:wrapTopAndBottom", &[]),
             Wrap::BehindText | Wrap::InFrontOfText => w.empty("wp:wrapNone", &[]),
         }
-        nv_props(w, "wp:docPr", docpr, name, alt, &float.alt, float.hidden);
+        nv_props(w, "wp:docPr", (docpr, name), alt, &float.alt, float.hidden, link);
     }
 }
 
@@ -502,8 +513,9 @@ pub(in super::super) fn spin_attrs(spin: wordcraft_geom::Spin) -> Vec<(&'static
 /// A drawing's `wp:docPr` (or a group member's `pic:cNvPr`/`wps:cNvPr`, ECMA-376
 /// `CT_NonVisualDrawingProps`) with its alt text `descr`, its title, and the decorative mark as
 /// the `adec:decorative` extension ([MS-ODRAWXML]). A description that is only the title (read
-/// from a file with no description) is not written twice.
-pub(in super::super) fn nv_props(w: &mut W, tag: &str, id: &str, name: &str, descr: &str, alt: &AltText, hidden: bool) {
+/// from a file with no description) is not written twice. `link` is the relationship id of the
+/// drawing's hyperlink (`a:hlinkClick`, the first child the schema allows).
+pub(in super::super) fn nv_props(w: &mut W, tag: &str, (id, name): (&str, &str), descr: &str, alt: &AltText, hidden: bool, link: Option<&str>) {
     let mut attrs = vec![("id", id), ("name", name)];
     if descr != alt.title || descr.is_empty() {
         attrs.push(("descr", descr));
@@ -514,16 +526,21 @@ pub(in super::super) fn nv_props(w: &mut W, tag: &str, id: &str, name: &str, des
     if hidden {
         attrs.push(("hidden", "1"));
     }
-    if !alt.decorative {
+    if !alt.decorative && link.is_none() {
         w.empty(tag, &attrs);
         return;
     }
     w.open(tag, &attrs);
-    w.open("a:extLst", &[]);
-    w.open("a:ext", &[("uri", crate::xml::DECORATIVE_EXT)]);
-    w.empty("adec:decorative", &[("xmlns:adec", crate::xml::DECORATIVE_NS), ("val", "1")]);
-    w.close("a:ext");
-    w.close("a:extLst");
+    if let Some(rid) = link {
+        w.empty("a:hlinkClick", &[("r:id", rid)]);
+    }
+    if alt.decorative {
+        w.open("a:extLst", &[]);
+        w.open("a:ext", &[("uri", crate::xml::DECORATIVE_EXT)]);
+        w.empty("adec:decorative", &[("xmlns:adec", crate::xml::DECORATIVE_NS), ("val", "1")]);
+        w.close("a:ext");
+        w.close("a:extLst");
+    }
     w.close(tag);
 }
 

crates/docx/src/write/story/object.rs

@@ -220,13 +220,18 @@ impl Writer<'_> {
                 w.open("mc:Choice", &[("Requires", requires)]);
             }
             let docpr = self.next_docpr();
+            let link = Self::drawing_link(props, rels);
             let name = match graphic.kind {
                 wordcraft_doc::graphic::GraphicKind::Chart => format!("Chart {docpr}"),
                 wordcraft_doc::graphic::GraphicKind::Diagram => format!("Diagram {docpr}"),
                 wordcraft_doc::graphic::GraphicKind::Ink => format!("Ink {docpr}"),
             };
             let z = self.z;
-            self.graphic_frame(w, DrawingInfo { float, size: (*gw, *gh), docpr: &docpr, name: &name, alt }, native.unwrap_or(&root));
+            self.graphic_frame(
+                w,
+                DrawingInfo { float, size: (*gw, *gh), docpr: &docpr, name: &name, alt, link: link.as_deref() },
+                native.unwrap_or(&root),
+            );
             if let Some(fallback) = fallback {
                 w.close("mc:Choice");
                 match picture {
@@ -234,7 +239,11 @@ impl Writer<'_> {
                         // Both branches are one object: same id and stacking order.
                         self.z = z;
                         w.open("mc:Fallback", &[]);
-                        self.graphic_frame(w, DrawingInfo { float, size: (*gw, *gh), docpr: &docpr, name: &name, alt }, picture);
+                        self.graphic_frame(
+                            w,
+                            DrawingInfo { float, size: (*gw, *gh), docpr: &docpr, name: &name, alt, link: link.as_deref() },
+                            picture,
+                        );
                         w.close("mc:Fallback");
                     }
                     None => w.raw(&fallback.to_xml()),
@@ -266,8 +275,9 @@ impl Writer<'_> {
         self.rpr(w, props);
         w.open("w:drawing", &[]);
         let docpr = self.next_docpr();
+        let link = Self::drawing_link(props, rels);
         let name = format!("Picture {docpr}");
-        self.drawing_open(w, DrawingInfo { float, size: (*iw, *ih), docpr: &docpr, name: &name, alt });
+        self.drawing_open(w, DrawingInfo { float, size: (*iw, *ih), docpr: &docpr, name: &name, alt, link: link.as_deref() });
         w.empty("wp:cNvGraphicFramePr", &[]);
         w.open("a:graphic", &[]);
         w.open("a:graphicData", &[("uri", "http://schemas.openxmlformats.org/drawingml/2006/picture")]);
@@ -355,6 +365,7 @@ impl Writer<'_> {
         self.rpr(w, props);
         w.open("w:drawing", &[]);
         let docpr = self.shape_docpr(extra.id);
+        let link = Self::drawing_link(props, rels);
         // Ink is told apart by its name (and its pen), which reading looks for.
         let name = shape_name(*kind, freeform.as_deref(), &docpr);
         // The effect extent leaves room for the shadow and glow.
@@ -369,7 +380,7 @@ impl Writer<'_> {
                 .map(std::sync::Arc::new);
         }
         let float = &float;
-        self.drawing_open(w, DrawingInfo { float, size: (*sw, *sh), docpr: &docpr, name: &name, alt: o.alt_text() });
+        self.drawing_open(w, DrawingInfo { float, size: (*sw, *sh), docpr: &docpr, name: &name, alt: o.alt_text(), link: link.as_deref() });
         w.empty("wp:cNvGraphicFramePr", &[]);
         w.open("a:graphic", &[]);
         w.open("a:graphicData", &[("uri", "http://schemas.microsoft.com/office/word/2010/wordprocessingShape")]);
@@ -389,8 +400,9 @@ impl Writer<'_> {
         self.rpr(w, props);
         w.open("w:drawing", &[]);
         let docpr = self.next_docpr();
+        let link = Self::drawing_link(props, rels);
         let name = format!("{} {docpr}", if canvas.is_some() { "Canvas" } else { "Group" });
-        self.drawing_open(w, DrawingInfo { float, size: (*gw, *gh), docpr: &docpr, name: &name, alt: o.alt_text() });
+        self.drawing_open(w, DrawingInfo { float, size: (*gw, *gh), docpr: &docpr, name: &name, alt: o.alt_text(), link: link.as_deref() });
         w.empty("wp:cNvGraphicFramePr", &[]);
         w.open("a:graphic", &[]);
         let outer = match canvas {

crates/docx/src/xml.rs

@@ -94,6 +94,49 @@ fn prefix_for(uri: &[u8], extra: &[(&'static str, &str)]) -> &'static str {
     "?"
 }
 
+/// Word's 2006 namespace (`wne`). A Strict document from Word keeps a text box's content in
+/// `wne:txbxContent`; a Transitional one uses `w:txbxContent`.
+const WNE: &[u8] = b"http://schemas.microsoft.com/office/word/2006/wordml";
+
+/// The canonical prefix of an element named `local` in namespace `uri`, inside an element
+/// named `parent` (canonical, `None` at the root). This is [`prefix_for`] with two exceptions
+/// for Strict documents (#3548):
+///
+/// - Shapes, groups and canvases (`wsp`, `wgp`, `wpc`) and their parts are elements of the
+///   WordprocessingDrawing namespace in ECMA-376 Part 1, 5th edition (`dml-wordprocessingDrawing.xsd`,
+///   Strict and Transitional alike). A Strict document from Word puts them there; a Transitional
+///   one uses Word 2010's own namespaces, whose names the reader knows. So they read as `wps:`,
+///   `wpg:` and `wpc:` (see [`drawing_prefix`]).
+/// - A shape's text box content reads as `w:txbxContent` from the drawing namespace (the
+///   schema's name) and from Word's `wne:txbxContent`.
+fn element_prefix(uri: &[u8], local: &str, parent: Option<&str>, extra: &[(&'static str, &str)]) -> &'static str {
+    match prefix_for(uri, extra) {
+        "wp" => drawing_prefix(local, parent),
+        "?" if uri == WNE && local == "txbxContent" && parent == Some("wps:txbx") => "w",
+        p => p,
+    }
+}
+
+/// The prefix for `local`, an element of the WordprocessingDrawing namespace inside `parent`.
+/// Word 2010's namespaces split what this one holds: a shape and its parts are `wps:`, a group
+/// (`wgp`, a nested `grpSp`) and a graphic frame in it are `wpg:`, and a canvas is `wpc:`. Its
+/// members are named for what they are. A text box's content is `w:txbxContent`, and anything
+/// else (`inline`, `anchor`, `docPr`…) stays `wp:`.
+fn drawing_prefix(local: &str, parent: Option<&str>) -> &'static str {
+    match (parent, local) {
+        (_, "wsp") => "wps",
+        (_, "wgp" | "grpSp" | "graphicFrame") => "wpg",
+        (_, "wpc") => "wpc",
+        (Some("wps:wsp"), "cNvPr" | "cNvSpPr" | "cNvCnPr" | "spPr" | "style" | "extLst" | "txbx" | "linkedTxbx" | "bodyPr") => "wps",
+        (Some("wps:txbx"), "txbxContent") => "w",
+        (Some("wps:txbx" | "wps:linkedTxbx"), "extLst") => "wps",
+        (Some("wpg:wgp" | "wpg:grpSp"), "cNvPr" | "cNvGrpSpPr" | "grpSpPr" | "extLst") => "wpg",
+        (Some("wpg:graphicFrame"), "cNvPr" | "cNvFrPr" | "xfrm" | "extLst") => "wpg",
+        (Some("wpc:wpc"), "bg" | "whole" | "extLst") => "wpc",
+        _ => "wp",
+    }
+}
+
 /// Compatibility feature namespaces whose elements are not otherwise handled by this tree.
 /// Keep the reader's existing feature set when canonicalizing `mc:Choice/@Requires` aliases.
 fn requirement_prefix(uri: &[u8]) -> &'static str {
@@ -342,8 +385,8 @@ fn parse_with(bytes: &[u8], extra: &[(&'static str, &str)], strict: bool) -> Res
     loop {
         buf.clear();
         let (res, ev) = r.read_resolved_event_into(&mut buf).map_err(|e| DocxError::Xml(e.to_string()))?;
-        let el_prefix = match &res {
-            ResolveResult::Bound(ns) => Some(prefix_for(ns.as_ref(), extra)),
+        let el_ns = match &res {
+            ResolveResult::Bound(ns) => Some(ns.as_ref()),
             _ => None,
         };
         match ev {
@@ -356,8 +399,8 @@ fn parse_with(bytes: &[u8], extra: &[(&'static str, &str)], strict: bool) -> Res
                     return Err(DocxError::Limit("too many XML elements".into()));
                 }
                 let local = String::from_utf8_lossy(e.local_name().as_ref()).into_owned();
-                let name = match el_prefix {
-                    Some(p) => format!("{p}:{local}"),
+                let name = match el_ns {
+                    Some(ns) => format!("{}:{local}", element_prefix(ns, &local, stack.last().map(|p| p.name.as_str()), extra)),
                     None => local,
                 };
                 let mut el = El { name, attrs: Vec::new(), kids: Vec::new() };
@@ -419,15 +462,22 @@ fn parse_with(bytes: &[u8], extra: &[(&'static str, &str)], strict: bool) -> Res
                     attach(&mut stack, &mut root, el);
                 }
             }
+            // Line ends are normalised as XML 1.0 says (§2.11): CR LF and a lone CR are read as
+            // LF, so a file saved with Windows line ends has no CR in its text. A CR written as
+            // `&#13;` is a character reference and stays.
             Event::Text(t) => {
                 if let Some(top) = stack.last_mut() {
-                    let s = t.decode().map_err(|e| DocxError::Xml(e.to_string()))?;
+                    let s = t.xml_content(XmlVersion::Explicit1_0).map_err(|e| DocxError::Xml(e.to_string()))?;
                     push_text(top, &s);
                 }
             }
             Event::CData(t) => {
                 if let Some(top) = stack.last_mut() {
-                    let s = String::from_utf8_lossy(&t).into_owned();
+                    // CDATA that isn't UTF-8 is decoded lossily, with its line ends normalised the same way.
+                    let s = t
+                        .xml_content(XmlVersion::Explicit1_0)
+                        .map(|s| s.into_owned())
+                        .unwrap_or_else(|_| String::from_utf8_lossy(&t).replace("\r\n", "\n").replace('\r', "\n"));
                     push_text(top, &s);
                 }
             }
@@ -593,6 +643,42 @@ mod tests {
         assert!(parse(b"<a></a><b></b>").is_ok(), "lenient parsing still stops at the root");
     }
 
+    /// Every element name in `e`, depth first.
+    fn names(e: &El, out: &mut Vec<String>) {
+        out.push(e.name.clone());
+        for c in e.els() {
+            names(c, out);
+        }
+    }
+
+    #[test]
+    fn strict_drawing_names_read_as_word_2010_names() {
+        // A Strict shape with a text box, a group with a nested group and a graphic frame, and
+        // a canvas, each in the one Strict WordprocessingDrawing namespace (#3548).
+        let xml = br#"<w:body xmlns:w="http://purl.oclc.org/ooxml/wordprocessingml/main" xmlns:a="http://purl.oclc.org/ooxml/drawingml/main" xmlns:d="http://purl.oclc.org/ooxml/drawingml/wordprocessingDrawing" xmlns:wne="http://schemas.microsoft.com/office/word/2006/wordml">
+<d:anchor><d:extent/><d:docPr/><d:cNvGraphicFramePr/><a:graphic><a:graphicData>
+<d:wsp><d:cNvPr/><d:cNvSpPr/><d:spPr/><d:style/><d:extLst/><d:txbx><wne:txbxContent><w:p/></wne:txbxContent></d:txbx><d:bodyPr/></d:wsp>
+<d:wgp><d:cNvPr/><d:cNvGrpSpPr/><d:grpSpPr/><d:grpSp><d:cNvGrpSpPr/><d:grpSpPr/><d:wsp><d:cNvCnPr/><d:spPr/><d:linkedTxbx/><d:bodyPr/></d:wsp></d:grpSp>
+<d:graphicFrame><d:cNvPr/><d:cNvFrPr/><d:xfrm/><a:graphic/><d:extLst/></d:graphicFrame><d:extLst/></d:wgp>
+<d:wpc><d:bg/><d:whole/><d:wsp><d:spPr/><d:txbx><d:txbxContent/><d:extLst/></d:txbx></d:wsp><d:wgp><d:grpSpPr/></d:wgp><d:extLst/></d:wpc>
+</a:graphicData></a:graphic></d:anchor>
+<wne:txbxContent/><d:inline><d:docPr/><d:bodyPr/><d:bg/></d:inline></w:body>"#;
+        let mut got = Vec::new();
+        names(&parse(xml).unwrap(), &mut got);
+        let want = "w:body wp:anchor wp:extent wp:docPr wp:cNvGraphicFramePr a:graphic a:graphicData \
+            wps:wsp wps:cNvPr wps:cNvSpPr wps:spPr wps:style wps:extLst wps:txbx w:txbxContent w:p wps:bodyPr \
+            wpg:wgp wpg:cNvPr wpg:cNvGrpSpPr wpg:grpSpPr wpg:grpSp wpg:cNvGrpSpPr wpg:grpSpPr wps:wsp wps:cNvCnPr wps:spPr wps:linkedTxbx wps:bodyPr \
+            wpg:graphicFrame wpg:cNvPr wpg:cNvFrPr wpg:xfrm a:graphic wpg:extLst wpg:extLst \
+            wpc:wpc wpc:bg wpc:whole wps:wsp wps:spPr wps:txbx w:txbxContent wps:extLst wpg:wgp wpg:grpSpPr wpc:extLst \
+            ?:txbxContent wp:inline wp:docPr wp:bodyPr wp:bg";
+        assert_eq!(got.join(" "), want);
+        // The Transitional namespaces read as before.
+        let xml = br#"<w:body xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main" xmlns:wp="http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing" xmlns:wps="http://schemas.microsoft.com/office/word/2010/wordprocessingShape" xmlns:wne="http://schemas.microsoft.com/office/word/2006/wordml"><wp:inline><wp:docPr/></wp:inline><wps:wsp><wps:txbx><w:txbxContent/></wps:txbx></wps:wsp><wne:toolbars/></w:body>"#;
+        let mut got = Vec::new();
+        names(&parse(xml).unwrap(), &mut got);
+        assert_eq!(got.join(" "), "w:body wp:inline wp:docPr wps:wsp wps:txbx w:txbxContent ?:toolbars");
+    }
+
     #[test]
     fn a_relationship_attribute_is_the_namespaced_one_or_the_alias_of_one() {
         // The relationship namespace, and VML's `o:relid`.
@@ -645,6 +731,28 @@ mod tests {
         assert_eq!(t.attr("xml:space"), Some("preserve"));
     }
 
+    /// CDATA that isn't UTF-8 is decoded lossily, and its line ends are still normalised.
+    #[test]
+    fn line_ends_in_cdata_that_isnt_utf8_are_normalised() {
+        let x =
+            b"<x:document xmlns:x=\"http://schemas.openxmlformats.org/wordprocessingml/2006/main\"><x:t><![CDATA[a\xff\r\nb\rc]]></x:t></x:document>";
+        assert_eq!(parse(x).unwrap().child("w:t").unwrap().text(), "a\u{FFFD}\nb\nc");
+    }
+
+    /// XML 1.0 §2.11: CR LF and a lone CR in text and CDATA are read as LF; `&#13;` stays a CR.
+    #[test]
+    fn line_ends_in_text_are_normalised_but_a_cr_reference_stays() {
+        let x = b"<x:document xmlns:x=\"http://schemas.openxmlformats.org/wordprocessingml/2006/main\"><x:t>a\r\nb\rc&#13;d<![CDATA[e\r\nf]]></x:t></x:document>";
+        let e = parse(x).unwrap();
+        assert_eq!(e.child("w:t").unwrap().text(), "a\nb\nc\rde\nf");
+        // The writer escapes a CR, so a CR in the text survives a save.
+        let x = format!(
+            "<x:document xmlns:x=\"http://schemas.openxmlformats.org/wordprocessingml/2006/main\"><x:t>{}</x:t></x:document>",
+            esc("a\r\nb\rc")
+        );
+        assert_eq!(parse(x.as_bytes()).unwrap().child("w:t").unwrap().text(), "a\r\nb\rc");
+    }
+
     #[test]
     fn choice_requirements_resolve_namespace_aliases_and_rebindings() {
         let x = br#"<mc:AlternateContent xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006" xmlns:pen="http://schemas.microsoft.com/office/word/2010/wordprocessingInk"><mc:Choice Requires="pen"/><mc:Choice xmlns:pen="urn:future" Requires="pen"/><mc:Choice Requires="undeclared"/></mc:AlternateContent>"#;

crates/docx/tests/it/content_controls.rs

@@ -342,3 +342,57 @@ fn toc_gallery_around_a_toc_whose_heading_is_in_a_block_control_round_trips() {
         assert!(xml.rfind("</w:sdt>").unwrap() < xml.rfind(">Chapter One<").unwrap(), "pass {pass}: {xml}");
     }
 }
+
+#[test]
+fn word_toc_gallery_with_field_end_after_entries_round_trips_without_update() {
+    let style = |s: &str, inner: &str| format!(r#"<w:p><w:pPr><w:pStyle w:val="{s}"/></w:pPr>{inner}</w:p>"#);
+    let begin = r#"<w:r><w:fldChar w:fldCharType="begin"/></w:r><w:r><w:instrText xml:space="preserve"> TOC \o "1-3" \h \z \u </w:instrText></w:r><w:r><w:fldChar w:fldCharType="separate"/></w:r>"#;
+    let content = format!(
+        "{}{}{}<w:p><w:r><w:fldChar w:fldCharType=\"end\"/></w:r></w:p>",
+        style("TOCHeading", &t("Contents")),
+        style("TOC1", &format!("{begin}{}", t("Chapter One\t1"))),
+        style("TOC2", &t("Section\t2"))
+    );
+    let pr = r#"<w:id w:val="-1234567"/><w:docPartObj><w:docPartGallery w:val="Table of Contents"/><w:docPartUnique/></w:docPartObj>"#;
+    let mut bytes = package(&format!("{}{}", sdt(pr, &content), p(&t("Chapter One"))));
+    for pass in 0..2 {
+        let d = read(&bytes).unwrap();
+        bytes = write(&d).unwrap();
+        let xml = document_xml(&bytes);
+        assert_eq!(xml.matches("<w:sdt>").count(), 1, "pass {pass}: {xml}");
+        assert!(xml.contains(r#"<w:docPartGallery w:val="Table of Contents"/>"#), "pass {pass}: {xml}");
+        assert!(xml.find("</w:sdt>").unwrap() < xml.rfind(">Chapter One<").unwrap(), "pass {pass}: {xml}");
+    }
+}
+
+#[test]
+fn range_toc_without_control_stays_unwrapped_and_section_break_toc_control_hands_over() {
+    let style = |s: &str, inner: &str| format!(r#"<w:p><w:pPr><w:pStyle w:val="{s}"/></w:pPr>{inner}</w:p>"#);
+    let begin = r#"<w:r><w:fldChar w:fldCharType="begin"/></w:r><w:r><w:instrText xml:space="preserve"> TOC \o "1-3" \h \z \u </w:instrText></w:r><w:r><w:fldChar w:fldCharType="separate"/></w:r>"#;
+    // A range TOC that was never in a content control (with FieldEnd in its own paragraph) must
+    // not gain a Table of Contents w:sdt on save.
+    let bare = format!(
+        "{}{}{}<w:p><w:r><w:fldChar w:fldCharType=\"end\"/></w:r></w:p>{}",
+        style("TOCHeading", &t("Contents")),
+        style("TOC1", &format!("{begin}{}", t("Chapter One\t1"))),
+        style("TOC2", &t("Section\t2")),
+        p(&t("Chapter One")),
+    );
+    let d_bare = read(&package(&bare)).unwrap();
+    let xml_bare = document_xml(&write(&d_bare).unwrap());
+    assert_eq!(xml_bare.matches("<w:sdt>").count(), 0, "{xml_bare}");
+
+    // #3889: a Table of Contents control opening on an empty paragraph that ends a section above
+    // the entries hands its control over on read so Update Table reuses the section-break paragraph.
+    let sect_toc = format!(
+        r#"<w:p><w:pPr><w:sectPr><w:pgSz w:w="11906" w:h="16838"/></w:sectPr></w:pPr></w:p>{}{}<w:p><w:r><w:fldChar w:fldCharType="end"/></w:r></w:p>"#,
+        style("TOC1", &format!("{begin}{}", t("Chapter One\t1"))),
+        style("TOC2", &t("Section\t2")),
+    );
+    let pr = r#"<w:id w:val="-1234567"/><w:docPartObj><w:docPartGallery w:val="Table of Contents"/><w:docPartUnique/></w:docPartObj>"#;
+    let d_sect = read(&package(&format!("{}{}", sdt(pr, &sect_toc), p(&t("Chapter One"))))).unwrap();
+    assert!(controls(&d_sect).is_empty());
+    assert!(d_sect.body[0].as_para().unwrap().section.is_some());
+    let xml_sect = document_xml(&write(&d_sect).unwrap());
+    assert!(xml_sect.contains("Chapter One") && xml_sect.contains("Section"), "{xml_sect}");
+}

crates/docx/tests/it/fixtures.rs

@@ -1766,12 +1766,27 @@ const CT_CHARTEX: &str = "application/vnd.ms-office.chartex+xml";
 /// A package holding one box & whisker chart the way Word writes it: the `cx:chart` in an
 /// `mc:Choice` requiring `requires` (bound to `requires_ns`), a picture of it in the `mc:Fallback`.
 fn chartex_docx(requires: &str, requires_ns: &str) -> Vec<u8> {
-    let frame = r#"<wp:inline distT="0" distB="0" distL="0" distR="0"><wp:extent cx="2743200" cy="1371600"/><wp:docPr id="1" name="Chart 1"/><wp:cNvGraphicFramePr/>"#;
+    chartex_docx_linked(requires, requires_ns, false)
+}
+
+/// [`chartex_docx`], its drawing linked to a web address (`wp:docPr/a:hlinkClick`) when `link`,
+/// after a first choice linked elsewhere, which needs a namespace no reader knows, so a reader
+/// takes the chart's choice.
+fn chartex_docx_linked(requires: &str, requires_ns: &str, link: bool) -> Vec<u8> {
+    let click = if link { r#"<a:hlinkClick r:id="rIdWeb"/>"# } else { "" };
+    let other = if link {
+        r#"<mc:Choice xmlns:w99="urn:example:unsupported" Requires="w99"><w:drawing><wp:inline distT="0" distB="0" distL="0" distR="0"><wp:extent cx="254000" cy="254000"/><wp:docPr id="9" name="Other"><a:hlinkClick r:id="rIdOther"/></wp:docPr><a:graphic><a:graphicData uri="http://schemas.microsoft.com/office/word/2010/wordprocessingShape"><wps:wsp><wps:spPr><a:prstGeom prst="rect"><a:avLst/></a:prstGeom></wps:spPr><wps:bodyPr/></wps:wsp></a:graphicData></a:graphic></wp:inline></w:drawing></mc:Choice>"#
+    } else {
+        ""
+    };
+    let frame = format!(
+        r#"<wp:inline distT="0" distB="0" distL="0" distR="0"><wp:extent cx="2743200" cy="1371600"/><wp:docPr id="1" name="Chart 1">{click}</wp:docPr><wp:cNvGraphicFramePr/>"#
+    );
     let body = format!(
-        r#"<w:p><w:r><w:t>Spread</w:t></w:r><w:r><mc:AlternateContent><mc:Choice xmlns:{requires}="{requires_ns}" Requires="{requires}"><w:drawing>{frame}<a:graphic><a:graphicData uri="{CHARTEX_NS}"><cx:chart xmlns:cx="{CHARTEX_NS}" r:id="rIdCx"/></a:graphicData></a:graphic></wp:inline></w:drawing></mc:Choice><mc:Fallback><w:drawing>{frame}<a:graphic><a:graphicData uri="http://schemas.openxmlformats.org/drawingml/2006/picture"><pic:pic><pic:nvPicPr><pic:cNvPr id="1" name="Chart 1"/><pic:cNvPicPr/></pic:nvPicPr><pic:blipFill><a:blip r:embed="rIdPic"/><a:stretch><a:fillRect/></a:stretch></pic:blipFill><pic:spPr><a:xfrm><a:off x="0" y="0"/><a:ext cx="2743200" cy="1371600"/></a:xfrm><a:prstGeom prst="rect"><a:avLst/></a:prstGeom></pic:spPr></pic:pic></a:graphicData></a:graphic></wp:inline></w:drawing></mc:Fallback></mc:AlternateContent></w:r></w:p>"#
+        r#"<w:p><w:r><w:t>Spread</w:t></w:r><w:r><mc:AlternateContent>{other}<mc:Choice xmlns:{requires}="{requires_ns}" Requires="{requires}"><w:drawing>{frame}<a:graphic><a:graphicData uri="{CHARTEX_NS}"><cx:chart xmlns:cx="{CHARTEX_NS}" r:id="rIdCx"/></a:graphicData></a:graphic></wp:inline></w:drawing></mc:Choice><mc:Fallback><w:drawing>{frame}<a:graphic><a:graphicData uri="http://schemas.openxmlformats.org/drawingml/2006/picture"><pic:pic><pic:nvPicPr><pic:cNvPr id="1" name="Chart 1"/><pic:cNvPicPr/></pic:nvPicPr><pic:blipFill><a:blip r:embed="rIdPic"/><a:stretch><a:fillRect/></a:stretch></pic:blipFill><pic:spPr><a:xfrm><a:off x="0" y="0"/><a:ext cx="2743200" cy="1371600"/></a:xfrm><a:prstGeom prst="rect"><a:avLst/></a:prstGeom></pic:spPr></pic:pic></a:graphicData></a:graphic></wp:inline></w:drawing></mc:Fallback></mc:AlternateContent></w:r></w:p>"#
     );
     let doc = format!(r#"<?xml version="1.0" encoding="UTF-8" standalone="yes"?><w:document {W_NS}><w:body>{body}</w:body></w:document>"#);
-    let doc_rels = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rIdCx" Type="http://schemas.microsoft.com/office/2014/relationships/chartEx" Target="charts/chartEx1.xml"/><Relationship Id="rIdPic" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/image" Target="media/image1.png"/></Relationships>"#;
+    let doc_rels = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rIdCx" Type="http://schemas.microsoft.com/office/2014/relationships/chartEx" Target="charts/chartEx1.xml"/><Relationship Id="rIdPic" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/image" Target="media/image1.png"/><Relationship Id="rIdWeb" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/hyperlink" Target="https://example.com/chart" TargetMode="External"/><Relationship Id="rIdOther" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/hyperlink" Target="https://example.com/other" TargetMode="External"/></Relationships>"#;
     let chart_rels = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/package" Target="../embeddings/Microsoft_Excel_Worksheet.xlsx"/><Relationship Id="rId2" Type="http://schemas.microsoft.com/office/2011/relationships/chartColorStyle" Target="colors1.xml"/></Relationships>"#;
     let types = format!(
         r#"<Types xmlns="http://schemas.openxmlformats.org/package/2006/content-types"><Default Extension="rels" ContentType="application/vnd.openxmlformats-package.relationships+xml"/><Default Extension="xml" ContentType="application/xml"/><Default Extension="png" ContentType="image/png"/><Default Extension="xlsx" ContentType="{CT_XLSX}"/><Override PartName="/word/document.xml" ContentType="application/vnd.openxmlformats-officedocument.wordprocessingml.document.main+xml"/><Override PartName="/word/charts/chartEx1.xml" ContentType="{CT_CHARTEX}"/><Override PartName="/word/charts/colors1.xml" ContentType="application/vnd.ms-office.chartcolorstyle+xml"/></Types>"#
@@ -1827,6 +1842,25 @@ fn office_2016_chart_keeps_its_chart_part_and_fallback_through_saves() {
     }
 }
 
+/// #5267: a link on an Office 2016 chart's drawing is the chart's link (not an earlier choice's),
+/// and saving keeps it on both branches (the writer rebuilds their frames).
+#[test]
+fn a_link_on_an_office_2016_charts_drawing_is_kept() {
+    let mut d = wordcraft_docx::read(&chartex_docx_linked("cx1", "http://schemas.microsoft.com/office/drawing/2015/9/8/chartex", true)).unwrap();
+    for _ in 0..2 {
+        let p = d.body[0].as_para().unwrap();
+        let off = p.object_offsets().into_iter().find(|o| matches!(p.object_at(*o), Some(InlineObject::Graphic { .. }))).unwrap();
+        assert_eq!(p.props_of_char(off).link.as_deref(), Some("https://example.com/chart"));
+        let files = unzip(&wordcraft_docx::write(&d).unwrap());
+        let body = text(&files, "word/document.xml");
+        let ids = attr_values(&body, "a:hlinkClick", "r:id");
+        assert_eq!(ids.len(), 2, "on the chart and its fallback picture: {body}");
+        assert_eq!(rel(&text(&files, "word/_rels/document.xml.rels"), &ids[0]).1, "https://example.com/chart");
+        let entries: Vec<_> = files.iter().map(|(name, bytes)| (name.as_str(), bytes.as_slice())).collect();
+        d = wordcraft_docx::read(&zip(&entries)).unwrap();
+    }
+}
+
 #[test]
 fn resized_office_2016_chart_resizes_its_fallback_too() {
     let mut d = wordcraft_docx::read(&chartex_docx("cx1", "http://schemas.microsoft.com/office/drawing/2015/9/8/chartex")).unwrap();
@@ -2462,6 +2496,31 @@ fn east_asian_number_formats_read_label_and_round_trip() {
     assert_eq!(formats(&back), want);
 }
 
+/// #4985: a level with numFmt none and literal text (`RESPONSE:`) labels its paragraph with that
+/// text, as in Word, and keeps both through a save.
+#[test]
+fn a_none_level_s_literal_text_is_its_label() {
+    use wordcraft_doc::numbering::Counters;
+    use wordcraft_doc::section::NumFormat;
+    let numbering = format!(
+        r#"<w:numbering {W_NS}><w:abstractNum w:abstractNumId="0"><w:lvl w:ilvl="0"><w:start w:val="1"/><w:numFmt w:val="none"/><w:lvlText w:val="RESPONSE:"/><w:pPr><w:ind w:left="1440" w:hanging="1440"/></w:pPr></w:lvl></w:abstractNum><w:num w:numId="1"><w:abstractNumId w:val="0"/></w:num></w:numbering>"#
+    );
+    let body = r#"<w:p><w:pPr><w:numPr><w:ilvl w:val="0"/><w:numId w:val="1"/></w:numPr></w:pPr><w:r><w:t>X</w:t></w:r></w:p>"#;
+    let bytes = docx(body, &[("rId1", "numbering", "numbering.xml")], &[("word/numbering.xml", &numbering)]);
+    let label = |d: &Document| {
+        let n = paras(d).first().and_then(|p| p.props.numbering)?;
+        Counters::default().next_label(&d.numbering, n.num, n.level).map(|l| l.0)
+    };
+    let d = wordcraft_docx::read(&bytes).unwrap();
+    assert_eq!(label(&d).as_deref(), Some("RESPONSE:"));
+    let back = wordcraft_docx::read(&wordcraft_docx::write(&d).unwrap()).unwrap();
+    assert_eq!(
+        back.numbering.level(1, 0).map(|l| (l.format, l.text.as_str(), l.indent, l.hanging)),
+        Some((NumFormat::None, "RESPONSE:", 72.0, 72.0))
+    );
+    assert_eq!(label(&back).as_deref(), Some("RESPONSE:"));
+}
+
 /// #464: a bookmark whose end follows the document's final table keeps its end (it goes into
 /// the table's last cell) and is saved as a matched pair.
 #[test]
@@ -3888,6 +3947,33 @@ fn bidi_theme_slots_follow_the_theme_font_language() {
     }
 }
 
+/// A `w:rFonts` fallback list longer than the reader keeps is cut at the last complete family
+/// within the bound, so the family a real list would pick survives while a hostile list cannot
+/// cost its whole length in every piece of text drawn in it.
+#[test]
+fn a_hostile_fallback_list_keeps_its_leading_families() {
+    let candidates: Vec<String> = (0..500).map(|i| format!("MissingCandidateFont{i}")).collect();
+    let list = format!("{};Source Serif 4", candidates.join(";"));
+    let body = format!(r#"<w:p><w:r><w:rPr><w:rFonts w:ascii="{list}" w:hAnsi="{list}"/></w:rPr><w:t>x</w:t></w:r></w:p>"#);
+    let d = read_body(&body);
+    let c = &d.body[0].as_para().unwrap().runs[0].props;
+    let kept = c.font.as_deref().expect("the leading families are kept");
+    assert!(kept.len() <= 256, "{} bytes kept", kept.len());
+    assert!(kept.starts_with("MissingCandidateFont0"), "{kept}");
+    // Every kept entry is a whole family name: the cut lands on a `;`, never inside a name.
+    assert!(kept.split(';').all(|f| f.starts_with("MissingCandidateFont") || f.contains("Source Serif")), "{kept}");
+    assert!(!kept.ends_with(';'), "no dangling separator: {kept}");
+}
+
+/// A real LibreOffice fallback list is read whole: a few families, well under any bound.
+#[test]
+fn a_real_fallback_list_is_read_whole() {
+    let body = r#"<w:p><w:r><w:rPr><w:rFonts w:ascii="Liberation Sans;Arial" w:hAnsi="Liberation Sans;Arial"/></w:rPr><w:t>x</w:t></w:r></w:p>"#;
+    let d = read_body(body);
+    let c = &d.body[0].as_para().unwrap().runs[0].props;
+    assert_eq!(c.font.as_deref(), Some("Liberation Sans;Arial"));
+}
+
 /// The font table's family classes: Word draws a missing roman font in a serif face.
 #[test]
 fn roman_fonts_are_read_from_the_font_table() {
@@ -3897,6 +3983,15 @@ fn roman_fonts_are_read_from_the_font_table() {
     assert_eq!(d.settings.roman_fonts, ["Brand Serif"]);
 }
 
+/// Semicolon-separated roman font fallback lists in `fontTable.xml` register each candidate family.
+#[test]
+fn roman_fonts_split_fallback_lists() {
+    let fonts = r#"<?xml version="1.0" encoding="UTF-8" standalone="yes"?><w:fonts xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:font w:name="Liberation Serif;Times New Roman"><w:family w:val="roman"/></w:font></w:fonts>"#;
+    let rels = [("rId9", "http://schemas.openxmlformats.org/officeDocument/2006/relationships/fontTable", "fontTable.xml")];
+    let d = wordcraft_docx::read(&docx("<w:p/>", &rels, &[("word/fontTable.xml", fonts)])).unwrap();
+    assert_eq!(d.settings.roman_fonts, ["Liberation Serif", "Times New Roman"]);
+}
+
 /// `w:doNotUseHTMLParagraphAutoSpacing` in the settings' `w:compat`: paragraph spacing adds up.
 #[test]
 fn html_auto_spacing_off_is_read_and_kept() {
@@ -4295,6 +4390,54 @@ fn zero_effects_stay_zero_rotated() {
     }
 }
 
+/// A turned picture whose file keeps no room for the turn (no or a zero `wp:effectExtent`) is
+/// marked so its line keeps to its frame as in Word, and saves that way; one with the room keeps
+/// its rotated bounds.
+#[test]
+fn bare_turns_are_kept_through_a_save() {
+    let pic = |effects: &str| {
+        format!(
+            r#"<w:p><w:r><w:drawing><wp:inline><wp:extent cx="1524000" cy="762000"/>{effects}<wp:docPr id="1" name="Synthetic picture"/><a:graphic><a:graphicData uri="http://schemas.openxmlformats.org/drawingml/2006/picture"><pic:pic><pic:blipFill><a:blip r:embed="rId1"/></pic:blipFill><pic:spPr><a:xfrm rot="2700000"><a:off x="0" y="0"/><a:ext cx="1524000" cy="762000"/></a:xfrm></pic:spPr></pic:pic></a:graphicData></a:graphic></wp:inline></w:drawing></w:r></w:p>"#
+        )
+    };
+    let read =
+        |body: &str| wordcraft_docx::read(&docx(body, &[("rId1", "image", "media/marker.png")], &[("word/media/marker.png", "synthetic")])).unwrap();
+    let bare = |d: &Document| paras(d)[0].objects[0].frame().unwrap().2.bare_turn;
+    for effects in ["", r#"<wp:effectExtent l="0" t="0" r="0" b="0"/>"#] {
+        let d = read(&pic(effects));
+        assert!(bare(&d), "{effects:?}");
+        let out = wordcraft_docx::write(&d).unwrap();
+        let body = text(&unzip(&out), "word/document.xml");
+        assert_eq!(attr_values(&body, "wp:effectExtent", "t"), ["0"]);
+        assert!(bare(&wordcraft_docx::read(&out).unwrap()));
+    }
+    // 120 × 60 pt at 45°: Word's rotated-bounds room, 3.64 pt wider and 33.64 pt taller each side.
+    let d = read(&pic(r#"<wp:effectExtent l="46210" t="427210" r="46210" b="427210"/>"#));
+    assert!(!bare(&d));
+    let mut d = read(&pic(""));
+    let f = first_object(&mut d).float_mut().unwrap();
+    f.set_spin(wordcraft_geom::Spin::new(30.0, false, false));
+    assert!(!f.bare_turn, "turned here: the rotated bounds again");
+    // A shadow's room added later: saved, and so laid out, with the turn's room again.
+    let mut d = read(&pic(""));
+    let f = first_object(&mut d).float_mut().unwrap();
+    f.effect = [2.0; 4];
+    assert!(!f.is_bare_turn());
+    let out = wordcraft_docx::write(&d).unwrap();
+    assert_ne!(attr_values(&text(&unzip(&out), "word/document.xml"), "wp:effectExtent", "t"), ["0"]);
+    // Made floating: saved with the turn's room, so it opens again as an ordinary turn.
+    let mut d = read(&pic(""));
+    let f = first_object(&mut d).float_mut().unwrap();
+    f.wrap = wordcraft_doc::para::Wrap::Square;
+    assert!(!f.is_bare_turn());
+    let out = wordcraft_docx::write(&d).unwrap();
+    assert_ne!(attr_values(&text(&unzip(&out), "word/document.xml"), "wp:effectExtent", "t"), ["0"]);
+    assert!(!bare(&wordcraft_docx::read(&out).unwrap()));
+    // Only inline ones: a floating one keeps text clear of its rotated bounds as before.
+    let anchored = pic("").replace("<wp:inline>", "<wp:anchor>").replace("</wp:inline>", "</wp:anchor>");
+    assert!(!bare(&read(&anchored)));
+}
+
 /// The math namespace, declared on the math element itself (the fixture root doesn't).
 const M_NS: &str = r#"xmlns:m="http://schemas.openxmlformats.org/officeDocument/2006/math""#;
 

crates/docx/tests/it/main.rs

@@ -21,8 +21,10 @@ mod heading_levels;
 mod legacy_forms;
 mod ligatures;
 mod malformed;
+mod nested_block_controls;
 mod note_comments;
 mod passthrough;
+mod picture_links;
 mod positional_tabs;
 mod privacy;
 mod roundtrip;
@@ -34,9 +36,11 @@ mod section_page_number_type;
 mod section_text_direction;
 mod section_valign;
 mod settings_merge;
+mod strict_drawings;
 mod text_alignment;
 mod text_boxes;
 mod toc_ranges;
+mod twips_wrap;
 mod undefined_style;
 mod vml_shapes;
 mod word_xml;

crates/docx/tests/it/nested_block_controls.rs

@@ -0,0 +1,180 @@
+//! A body-level content control nested directly in another body-level one stays body-level when
+//! saved, even when it holds a single paragraph (ECMA-376 §17.5.2.29-30; [MS-DOCX] `w15`
+//! repeating sections, which must wrap whole paragraphs or rows). All markup here is synthetic.
+
+use std::io::Write;
+
+use wordcraft_doc::{ContentControl, Document, StoryRef};
+use wordcraft_docx::{read, write};
+
+const NS: &str =
+    r#"xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main" xmlns:w15="http://schemas.microsoft.com/office/word/2012/wordml""#;
+const ROOT_RELS: &str = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
+
+fn package(body: &str) -> Vec<u8> {
+    let doc = format!(r#"<w:document {NS}><w:body>{body}</w:body></w:document>"#);
+    let mut zw = zip::ZipWriter::new(std::io::Cursor::new(Vec::new()));
+    for (name, bytes) in [("_rels/.rels", ROOT_RELS.as_bytes()), ("word/document.xml", doc.as_bytes())] {
+        zw.start_file(name, zip::write::SimpleFileOptions::default()).unwrap();
+        zw.write_all(bytes).unwrap();
+    }
+    zw.finish().unwrap().into_inner()
+}
+
+fn document_xml(docx: &[u8]) -> String {
+    let mut z = zip::ZipArchive::new(std::io::Cursor::new(docx)).unwrap();
+    let mut s = String::new();
+    std::io::Read::read_to_string(&mut z.by_name("word/document.xml").unwrap(), &mut s).unwrap();
+    s
+}
+
+fn t(s: &str) -> String {
+    format!(r#"<w:r><w:t xml:space="preserve">{s}</w:t></w:r>"#)
+}
+fn p(inner: &str) -> String {
+    format!("<w:p>{inner}</w:p>")
+}
+fn sdt(pr: &str, content: &str) -> String {
+    format!("<w:sdt><w:sdtPr>{pr}</w:sdtPr><w:sdtContent>{content}</w:sdtContent></w:sdt>")
+}
+
+fn controls(d: &Document) -> Vec<ContentControl> {
+    let mut out = Vec::new();
+    wordcraft_doc::control::walk_controls(&d.body, &mut |c| out.push(c.clone()));
+    out
+}
+
+/// How many `w:p` elements are open at byte `at` of `xml`.
+fn open_paragraphs(xml: &str, at: usize) -> usize {
+    let head = &xml[..at];
+    let opened = head.matches("<w:p>").count() + head.matches("<w:p ").count();
+    opened - head.matches("</w:p>").count()
+}
+
+/// Paragraph depth at the start of the `w:sdt` whose `w:sdtPr` holds `marker`.
+fn depth_of_control(xml: &str, marker: &str) -> usize {
+    let at = xml.find(marker).unwrap_or_else(|| panic!("{marker} not written:\n{xml}"));
+    let start = xml[..at].rfind("<w:sdt>").expect("control start");
+    open_paragraphs(xml, start)
+}
+
+fn round_trip(body: &str) -> (Document, Document, String) {
+    let d = read(&package(body)).unwrap();
+    let bytes = write(&d).unwrap();
+    let back = read(&bytes).unwrap();
+    (d, back, document_xml(&bytes))
+}
+
+fn repeating_section() -> String {
+    let field = sdt(r#"<w:alias w:val="note"/><w:id w:val="3"/><w:text/>"#, &t("note 1"));
+    let item = sdt(r#"<w:alias w:val="item"/><w:id w:val="2"/><w15:repeatingSectionItem/>"#, &p(&format!("{field}{}", t("   "))));
+    let section = sdt(r#"<w:alias w:val="notes"/><w:id w:val="1"/><w15:repeatingSection/>"#, &item);
+    format!("{}{section}{}", p(&t("Notes")), p(&t("End")))
+}
+
+#[test]
+fn a_repeating_section_around_one_paragraph_stays_around_the_paragraph() {
+    let (_, _, xml) = round_trip(&repeating_section());
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="notes"/>"#), 0, "section written inside a paragraph:\n{xml}");
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="item"/>"#), 0, "item written inside a paragraph:\n{xml}");
+    // The text control stays inside the paragraph, and the item holds that paragraph.
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="note"/>"#), 1);
+    assert!(xml.contains("<w15:repeatingSectionItem/></w:sdtPr><w:sdtContent><w:p"), "{xml}");
+}
+
+#[test]
+fn a_repeating_section_round_trips_its_controls_and_text() {
+    let (d, back, _) = round_trip(&repeating_section());
+    assert_eq!(controls(&back), controls(&d));
+    assert_eq!(back.plain_text(StoryRef::Body), d.plain_text(StoryRef::Body));
+    assert_eq!(back.body.len(), d.body.len());
+}
+
+#[test]
+fn plain_nested_block_controls_around_one_paragraph_stay_block_level() {
+    let inner = sdt(r#"<w:alias w:val="inner"/><w:id w:val="11"/>"#, &p(&t("Only paragraph")));
+    let outer = sdt(r#"<w:alias w:val="outer"/><w:id w:val="10"/>"#, &inner);
+    let (d, back, xml) = round_trip(&format!("{outer}{}", p(&t("After"))));
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="outer"/>"#), 0, "{xml}");
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="inner"/>"#), 0, "{xml}");
+    assert_eq!(controls(&back), controls(&d));
+}
+
+#[test]
+fn nested_block_controls_in_a_table_cell_stay_block_level() {
+    let inner = sdt(r#"<w:alias w:val="cellInner"/><w:id w:val="21"/>"#, &p(&t("In cell")));
+    let outer = sdt(r#"<w:alias w:val="cellOuter"/><w:id w:val="20"/>"#, &inner);
+    let table = format!(r#"<w:tbl><w:tblGrid><w:gridCol w:w="4000"/></w:tblGrid><w:tr><w:tc>{outer}</w:tc></w:tr></w:tbl>"#);
+    let (_, _, xml) = round_trip(&format!("{table}{}", p(&t("After"))));
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="cellOuter"/>"#), 0, "{xml}");
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="cellInner"/>"#), 0, "{xml}");
+}
+
+#[test]
+fn a_run_level_control_stays_inside_its_paragraph() {
+    let field = sdt(r#"<w:alias w:val="inline"/><w:id w:val="30"/>"#, &t("value"));
+    let (_, _, xml) = round_trip(&p(&format!("{}{field}", t("Name: "))));
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="inline"/>"#), 1, "{xml}");
+}
+
+/// A bound repeating section as Word writes one at the end of a form, optionally followed by
+/// the body-level `_GoBack` bookmark Word leaves after the last control.
+fn bound_section_body(bound: bool, bookmark: bool) -> String {
+    let section_binding = if bound {
+        r#"<w15:dataBinding w:prefixMappings="" w:xpath="/a[1]/n[1]" w:storeItemID="{00000000-0000-0000-0000-0000000000AA}"/>"#
+    } else {
+        ""
+    };
+    let trailing = if bookmark { r#"<w:bookmarkStart w:id="0" w:name="_GoBack"/><w:bookmarkEnd w:id="0"/>"# } else { "" };
+    let bind = |x: &str| format!(r#"<w:dataBinding w:prefixMappings="" w:xpath="{x}" w:storeItemID="{{00000000-0000-0000-0000-0000000000AA}}"/>"#);
+    let field = format!(
+        r#"<w:sdt><w:sdtPr><w:alias w:val="note"/><w:placeholder><w:docPart w:val="DefaultPlaceholder_1"/></w:placeholder>{}<w:text/></w:sdtPr><w:sdtEndPr/><w:sdtContent>{}</w:sdtContent></w:sdt>"#,
+        if bound { bind("/a[1]/n[1]") } else { String::new() },
+        t("note 1")
+    );
+    let item = format!(
+        r#"<w:sdt><w:sdtPr><w:alias w:val="item"/><w:placeholder><w:docPart w:val="DefaultPlaceholder_2"/></w:placeholder><w15:repeatingSectionItem/></w:sdtPr><w:sdtEndPr/><w:sdtContent>{}</w:sdtContent></w:sdt>"#,
+        p(&format!("{field}{}", t("   ")))
+    );
+    let section = format!(
+        r#"<w:sdt><w:sdtPr><w:alias w:val="notes"/>{section_binding}<w15:repeatingSection/></w:sdtPr><w:sdtContent>{item}</w:sdtContent></w:sdt>"#
+    );
+    format!(r#"{}{section}{trailing}<w:sectPr><w:pgSz w:w="12240" w:h="15840"/></w:sectPr>"#, p(&t("Notes")))
+}
+
+fn assert_section_block_level(body: &str) {
+    let (d, back, xml) = round_trip(body);
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="notes"/>"#), 0, "section written inside a paragraph:\n{xml}");
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="item"/>"#), 0, "item written inside a paragraph:\n{xml}");
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="note"/>"#), 1, "{xml}");
+    assert_eq!(controls(&back), controls(&d));
+    assert_eq!(back.plain_text(StoryRef::Body), d.plain_text(StoryRef::Body));
+}
+
+#[test]
+fn a_bound_repeating_section_followed_by_a_body_bookmark_stays_block_level() {
+    assert_section_block_level(&bound_section_body(true, true));
+}
+
+#[test]
+fn a_bound_repeating_section_at_the_end_of_the_body_stays_block_level() {
+    assert_section_block_level(&bound_section_body(true, false));
+}
+
+#[test]
+fn a_repeating_section_followed_by_a_body_bookmark_stays_block_level() {
+    assert_section_block_level(&bound_section_body(false, true));
+}
+
+#[test]
+fn a_block_control_followed_by_a_body_bookmark_stays_around_its_paragraph() {
+    let inner = sdt(r#"<w:alias w:val="inner"/><w:id w:val="41"/>"#, &p(&t("Last paragraph")));
+    let outer = sdt(r#"<w:alias w:val="outer"/><w:id w:val="40"/>"#, &inner);
+    let body = format!(r#"{}{outer}<w:bookmarkStart w:id="0" w:name="_GoBack"/><w:bookmarkEnd w:id="0"/>"#, p(&t("First")));
+    let (d, back, xml) = round_trip(&body);
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="outer"/>"#), 0, "{xml}");
+    assert_eq!(depth_of_control(&xml, r#"<w:alias w:val="inner"/>"#), 0, "{xml}");
+    assert_eq!(controls(&back), controls(&d));
+    // The bookmark is kept.
+    assert!(xml.contains(r#"w:name="_GoBack""#), "{xml}");
+}

crates/docx/tests/it/picture_links.rs

@@ -0,0 +1,127 @@
+//! A picture's hyperlink (#5267): Word keeps it on the drawing (`wp:docPr/a:hlinkClick`,
+//! ECMA-376 Part 1 §20.4.2.5) rather than in a `w:hyperlink` around the run. Original
+//! synthetic packages; the 2×2 picture is generated in code.
+use std::io::{Read, Write};
+use std::sync::Arc;
+use wordcraft_doc::{CharProps, Document, InlineObject, Pos};
+
+fn png() -> Vec<u8> {
+    let mut out = Vec::new();
+    image::RgbaImage::from_pixel(2, 2, image::Rgba([20, 90, 200, 255]))
+        .write_to(&mut std::io::Cursor::new(&mut out), image::ImageFormat::Png)
+        .unwrap();
+    out
+}
+
+/// An inline picture's run; `link` goes into its `wp:docPr`.
+fn picture_run(id: u32, link: &str) -> String {
+    format!(
+        r#"<w:r><w:drawing><wp:inline distT="0" distB="0" distL="0" distR="0"><wp:extent cx="254000" cy="254000"/><wp:docPr id="{id}" name="Picture {id}">{link}</wp:docPr><a:graphic><a:graphicData uri="http://schemas.openxmlformats.org/drawingml/2006/picture"><pic:pic><pic:nvPicPr><pic:cNvPr id="0" name="square.png"/><pic:cNvPicPr/></pic:nvPicPr><pic:blipFill><a:blip r:embed="rIdPic"/></pic:blipFill><pic:spPr><a:xfrm><a:off x="0" y="0"/><a:ext cx="254000" cy="254000"/></a:xfrm><a:prstGeom prst="rect"/></pic:spPr></pic:pic></a:graphicData></a:graphic></wp:inline></w:drawing></w:r>"#
+    )
+}
+
+fn package(body: &str) -> Vec<u8> {
+    let xml = format!(
+        r#"<?xml version="1.0" encoding="UTF-8" standalone="yes"?><w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main" xmlns:r="http://schemas.openxmlformats.org/officeDocument/2006/relationships" xmlns:wp="http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing" xmlns:a="http://schemas.openxmlformats.org/drawingml/2006/main" xmlns:pic="http://schemas.openxmlformats.org/drawingml/2006/picture"><w:body>{body}<w:sectPr/></w:body></w:document>"#
+    );
+    let rels = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
+    let doc_rels = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rIdPic" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/image" Target="media/square.png"/><Relationship Id="rIdWeb" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/hyperlink" Target="https://example.com/picture" TargetMode="External"/><Relationship Id="rIdRun" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/hyperlink" Target="https://example.com/run" TargetMode="External"/></Relationships>"#;
+    let types = r#"<Types xmlns="http://schemas.openxmlformats.org/package/2006/content-types"><Default Extension="rels" ContentType="application/vnd.openxmlformats-package.relationships+xml"/><Default Extension="png" ContentType="image/png"/><Override PartName="/word/document.xml" ContentType="application/vnd.openxmlformats-officedocument.wordprocessingml.document.main+xml"/></Types>"#;
+    let png = png();
+    let mut zip = zip::ZipWriter::new(std::io::Cursor::new(Vec::new()));
+    for (name, bytes) in [
+        ("[Content_Types].xml", types.as_bytes()),
+        ("_rels/.rels", rels.as_bytes()),
+        ("word/document.xml", xml.as_bytes()),
+        ("word/_rels/document.xml.rels", doc_rels.as_bytes()),
+        ("word/media/square.png", png.as_slice()),
+    ] {
+        zip.start_file(name, zip::write::SimpleFileOptions::default()).unwrap();
+        zip.write_all(bytes).unwrap();
+    }
+    zip.finish().unwrap().into_inner()
+}
+
+/// The link on each picture of the first paragraph, in order.
+fn picture_links(doc: &Document) -> Vec<Option<String>> {
+    let p = doc.body[0].as_para().unwrap();
+    p.object_offsets()
+        .into_iter()
+        .filter(|off| matches!(p.object_at(*off), Some(InlineObject::Image { .. })))
+        .map(|off| p.props_of_char(off).link.clone())
+        .collect()
+}
+
+fn part(bytes: &[u8], name: &str) -> String {
+    let mut z = zip::ZipArchive::new(std::io::Cursor::new(bytes)).unwrap();
+    let mut s = String::new();
+    z.by_name(name).unwrap().read_to_string(&mut s).unwrap();
+    s
+}
+
+#[test]
+fn a_link_on_a_pictures_drawing_is_read() {
+    let body = format!(
+        "<w:p>{}{}{}<w:hyperlink r:id=\"rIdRun\">{}</w:hyperlink></w:p>",
+        // Word's way: the link on the drawing.
+        picture_run(1, r#"<a:hlinkClick r:id="rIdWeb"/>"#),
+        // A click naming a relationship that is no hyperlink, and one naming none.
+        picture_run(2, r#"<a:hlinkClick r:id="rIdPic"/>"#),
+        picture_run(3, r#"<a:hlinkClick r:id="rIdMissing"/>"#),
+        // A link around the run wins over the drawing's.
+        picture_run(4, r#"<a:hlinkClick r:id="rIdWeb"/>"#),
+    );
+    let doc = wordcraft_docx::read(&package(&body)).unwrap();
+    assert_eq!(picture_links(&doc), [Some("https://example.com/picture".to_string()), None, None, Some("https://example.com/run".to_string())]);
+}
+
+#[test]
+fn a_linked_picture_is_saved_with_its_link_on_the_drawing_too() {
+    let mut doc = Document::new();
+    doc.media.insert("square.png".into(), Arc::new(png()));
+    let picture = || InlineObject::Image {
+        media: "square.png".into(),
+        w: 20.0,
+        h: 20.0,
+        alt: String::new(),
+        float: Default::default(),
+        crop: [0.0; 4],
+        ole: None,
+        fx: Default::default(),
+    };
+    let linked = |l: &str| CharProps { link: Some(l.to_string()), ..Default::default() };
+    // A place in this document stays a `w:hyperlink w:anchor` only: `a:hlinkClick` names a relationship.
+    doc.insert_object(&Pos::body(0, 0), picture(), &linked("#Spot")).unwrap();
+    doc.insert_object(&Pos::body(0, 0), picture(), &linked("https://example.com/saved")).unwrap();
+    let bytes = wordcraft_docx::write(&doc).unwrap();
+    let xml = part(&bytes, "word/document.xml");
+    let clicks: Vec<&str> = xml.match_indices("<a:hlinkClick ").map(|(i, _)| &xml[i..]).collect();
+    assert_eq!(clicks.len(), 1, "{xml}");
+    let rid = clicks[0].split('"').nth(1).unwrap();
+    let docpr = xml.find("<wp:docPr ").unwrap();
+    assert!(docpr < xml.find("<a:hlinkClick ").unwrap() && xml.find("<a:hlinkClick ").unwrap() < xml.find("</wp:docPr>").unwrap(), "{xml}");
+    let rels = part(&bytes, "word/_rels/document.xml.rels");
+    assert!(rels.contains(&format!(r#"Id="{rid}""#)) && rels.contains(r#"Target="https://example.com/saved""#), "{rels}");
+    // The run's own hyperlink names the same relationship, so the address is stored once.
+    assert_eq!(rels.matches("https://example.com/saved").count(), 1, "{rels}");
+    let back = wordcraft_docx::read(&bytes).unwrap();
+    assert_eq!(picture_links(&back), [Some("https://example.com/saved".to_string()), Some("#Spot".to_string())]);
+    // The drawing's link alone (as Word writes it) reads back the same.
+    let without_run_link = xml.replace(&format!(r#"<w:hyperlink r:id="{rid}" w:history="1">"#), "").replacen("</w:hyperlink>", "", 1);
+    assert_ne!(without_run_link, xml);
+    let mut z = zip::ZipArchive::new(std::io::Cursor::new(bytes.as_slice())).unwrap();
+    let mut out = zip::ZipWriter::new(std::io::Cursor::new(Vec::new()));
+    for i in 0..z.len() {
+        let mut f = z.by_index(i).unwrap();
+        let name = f.name().to_string();
+        let mut data = Vec::new();
+        f.read_to_end(&mut data).unwrap();
+        if name == "word/document.xml" {
+            data = without_run_link.clone().into_bytes();
+        }
+        out.start_file(name, zip::write::SimpleFileOptions::default()).unwrap();
+        out.write_all(&data).unwrap();
+    }
+    let edited = out.finish().unwrap().into_inner();
+    assert_eq!(picture_links(&wordcraft_docx::read(&edited).unwrap())[0], Some("https://example.com/saved".to_string()));
+}

crates/docx/tests/it/roundtrip.rs

@@ -6,8 +6,8 @@ use wordcraft_doc::fill::Fill;
 use wordcraft_doc::numbering::ListKind;
 use wordcraft_doc::para::{Anchor, Float, FloatAlign, NoteKind, ShapeKind, Wrap};
 use wordcraft_doc::props::{
-    Align, Border, BorderStyle, Borders, CharProps, HeightRule, Highlight, LineSpacing, NumRef, ParaProps, Rgb, RowProps, TabAlign, TabLeader,
-    TabStop, TableLook, TextAlignment, TextColor, TextDirection, Underline, VAlign, VMerge, VertAlign,
+    Align, Border, BorderStyle, Borders, CharProps, HeightRule, Highlight, LineSpacing, NumRef, ParaProps, ParagraphTextDirection, Rgb, RowProps,
+    TabAlign, TabLeader, TabStop, TableLook, TextAlignment, TextColor, TextDirection, Underline, VAlign, VMerge, VertAlign,
 };
 use wordcraft_doc::section::{Columns, DocGrid, DocGridType, LineNumberRestart, LineNumbering, NumFormat, PageVAlign, SectionProps, SectionStart};
 use wordcraft_doc::styles::{Style, StyleKind};
@@ -195,6 +195,7 @@ fn every_para_prop_round_trips() {
         suppress_hyphens: Some(true),
         suppress_line_numbers: Some(true),
         bidi: Some(false),
+        text_direction: Some(ParagraphTextDirection::TbRl),
         drop_cap: None,
         drop_cap_margin: None,
         drop_cap_distance: None,

crates/docx/tests/it/section_text_direction.rs

@@ -18,6 +18,25 @@ fn docx(body: &str) -> Vec<u8> {
     zw.finish().unwrap().into_inner()
 }
 
+fn docx_with_styles(body: &str, styles: &str) -> Vec<u8> {
+    let mut bytes = zip::ZipWriter::new(std::io::Cursor::new(Vec::new()));
+    let root = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
+    let document_rels = r#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rIdStyles" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/styles" Target="styles.xml"/></Relationships>"#;
+    let doc = format!(
+        r#"<?xml version="1.0"?><w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:body>{body}</w:body></w:document>"#
+    );
+    for (name, value) in [
+        ("_rels/.rels", root.as_bytes()),
+        ("word/_rels/document.xml.rels", document_rels.as_bytes()),
+        ("word/document.xml", doc.as_bytes()),
+        ("word/styles.xml", styles.as_bytes()),
+    ] {
+        bytes.start_file(name, zip::write::SimpleFileOptions::default()).unwrap();
+        bytes.write_all(value).unwrap();
+    }
+    bytes.finish().unwrap().into_inner()
+}
+
 fn document_xml(bytes: &[u8]) -> String {
     let mut z = zip::ZipArchive::new(std::io::Cursor::new(bytes)).unwrap();
     let mut s = String::new();
@@ -59,3 +78,52 @@ fn explicit_horizontal_direction_is_kept_in_sections_and_cells() {
     let plain = wordcraft_docx::read(&docx(r#"<w:p><w:r><w:t>Body</w:t></w:r></w:p><w:sectPr/>"#)).unwrap();
     assert!(!document_xml(&wordcraft_docx::write(&plain).unwrap()).contains("textDirection"));
 }
+
+#[test]
+fn paragraph_text_direction_round_trips_through_two_saves_and_json() {
+    use wordcraft_doc::props::ParaProps;
+    use wordcraft_doc::props::ParagraphTextDirection;
+
+    let values = ["lrTb", "lrTbV", "tb", "tbRl", "tbRlV", "tbLr", "tbLrV", "btLr", "tbV", "rl", "rlV", "lr", "lrV"];
+    let mut body: String =
+        values.iter().map(|value| format!(r#"<w:p><w:pPr><w:textDirection w:val="{value}"/></w:pPr><w:r><w:t>{value}</w:t></w:r></w:p>"#)).collect();
+    body.push_str(r#"<w:p><w:r><w:t>Inherited</w:t></w:r></w:p>"#);
+    let d = wordcraft_docx::read(&docx(&body)).unwrap();
+    let paras: Vec<_> = d.body.iter().filter_map(|b| b.as_para()).collect();
+    for (para, value) in paras.iter().zip(values) {
+        assert_eq!(para.props.text_direction, ParagraphTextDirection::from_ooxml(value), "{value}");
+    }
+    assert_eq!(paras[values.len()].props.text_direction, None);
+    let json = serde_json::to_value(&paras[0].props).unwrap();
+    assert_eq!(json["textDirection"], "lrTb");
+    let decoded: ParaProps = serde_json::from_value(json).unwrap();
+    assert_eq!(decoded.text_direction, Some(ParagraphTextDirection::LrTb));
+
+    let first = wordcraft_docx::write(&d).unwrap();
+    let xml = document_xml(&first);
+    for value in values {
+        assert!(xml.contains(&format!(r#"<w:textDirection w:val="{value}"/>"#)), "{xml}");
+    }
+    assert_eq!(xml.matches("w:textDirection").count(), values.len(), "inherited direction stays absent: {xml}");
+    let second = wordcraft_docx::write(&wordcraft_docx::read(&first).unwrap()).unwrap();
+    let twice = document_xml(&second);
+    assert_eq!(twice.matches("w:textDirection").count(), values.len(), "{twice}");
+    for value in values {
+        assert!(twice.contains(&format!(r#"<w:textDirection w:val="{value}"/>"#)), "{twice}");
+    }
+}
+
+#[test]
+fn paragraph_style_text_direction_is_preserved_and_direct_value_overrides_it() {
+    let styles = r#"<w:styles xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:style w:type="paragraph" w:styleId="Vertical"><w:name w:val="Vertical"/><w:pPr><w:textDirection w:val="tbRl"/></w:pPr></w:style></w:styles>"#;
+    let body = r#"<w:p><w:pPr><w:pStyle w:val="Vertical"/></w:pPr><w:r><w:t>Style</w:t></w:r></w:p><w:p><w:pPr><w:pStyle w:val="Vertical"/><w:textDirection w:val="lrTb"/></w:pPr><w:r><w:t>Override</w:t></w:r></w:p>"#;
+    let d = wordcraft_docx::read(&docx_with_styles(body, styles)).unwrap();
+    let xml = document_xml(&wordcraft_docx::write(&d).unwrap());
+    assert_eq!(xml.matches(r#"<w:pStyle w:val="Vertical"/>"#).count(), 2, "both paragraphs keep their style: {xml}");
+    assert!(xml.contains(r#"<w:pStyle w:val="Vertical"/><w:textDirection w:val="lrTb"/>"#), "direct horizontal is explicit: {xml}");
+    assert_eq!(xml.matches("w:textDirection").count(), 1, "only the direct paragraph writes a direction: {xml}");
+    let mut z = zip::ZipArchive::new(std::io::Cursor::new(wordcraft_docx::write(&d).unwrap())).unwrap();
+    let mut styles_xml = String::new();
+    z.by_name("word/styles.xml").unwrap().read_to_string(&mut styles_xml).unwrap();
+    assert!(styles_xml.contains(r#"<w:textDirection w:val="tbRl"/>"#), "style value survives: {styles_xml}");
+}

crates/docx/tests/it/strict_drawings.rs

@@ -0,0 +1,182 @@
+//! Shapes, text boxes, groups and canvases in a Strict document (#3548). ISO/IEC 29500 Strict
+//! puts them in the one WordprocessingDrawing namespace, where Transitional documents use Word
+//! 2010's `wps`, `wpg` and `wpc` namespaces. Fixtures are original and hand-written from
+//! ECMA-376 Part 1, 5th edition (`dml-wordprocessingDrawing.xsd`, Strict and Transitional): each
+//! drawing is written once and filled in with either set of namespaces.
+
+use std::io::{Read, Write};
+
+use wordcraft_doc::{Document, InlineObject, PartKind};
+
+const STRICT_NS: &str = r#"xmlns:w="http://purl.oclc.org/ooxml/wordprocessingml/main" xmlns:r="http://purl.oclc.org/ooxml/officeDocument/relationships" xmlns:wp="http://purl.oclc.org/ooxml/drawingml/wordprocessingDrawing" xmlns:a="http://purl.oclc.org/ooxml/drawingml/main" xmlns:wne="http://schemas.microsoft.com/office/word/2006/wordml" w:conformance="strict""#;
+const TRANSITIONAL_NS: &str = r#"xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main" xmlns:r="http://schemas.openxmlformats.org/officeDocument/2006/relationships" xmlns:wp="http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing" xmlns:a="http://schemas.openxmlformats.org/drawingml/2006/main" xmlns:wps="http://schemas.microsoft.com/office/word/2010/wordprocessingShape" xmlns:wpg="http://schemas.microsoft.com/office/word/2010/wordprocessingGroup" xmlns:wpc="http://schemas.microsoft.com/office/word/2010/wordprocessingCanvas""#;
+const STRICT_RELS: &str = r#"<?xml version="1.0" encoding="UTF-8"?><Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://purl.oclc.org/ooxml/officeDocument/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
+const TRANSITIONAL_RELS: &str = r#"<?xml version="1.0" encoding="UTF-8"?><Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
+
+/// How a fixture names its drawing elements: the shape, group and canvas prefixes, and the
+/// text box content element.
+struct Names {
+    ns: &'static str,
+    rels: &'static str,
+    shape: &'static str,
+    group: &'static str,
+    canvas: &'static str,
+    content: &'static str,
+}
+
+/// Word's Strict output: everything in `wp:`, text box content in `wne:txbxContent`.
+const STRICT_WORD: Names = Names { ns: STRICT_NS, rels: STRICT_RELS, shape: "wp", group: "wp", canvas: "wp", content: "wne:txbxContent" };
+/// The Strict schema's own names: the text box content in `wp:` too.
+const STRICT_SCHEMA: Names = Names { ns: STRICT_NS, rels: STRICT_RELS, shape: "wp", group: "wp", canvas: "wp", content: "wp:txbxContent" };
+/// The same drawings in a Transitional document.
+const TRANSITIONAL: Names =
+    Names { ns: TRANSITIONAL_NS, rels: TRANSITIONAL_RELS, shape: "wps", group: "wpg", canvas: "wpc", content: "w:txbxContent" };
+
+fn zip(entries: &[(&str, &[u8])]) -> Vec<u8> {
+    let mut zw = zip::ZipWriter::new(std::io::Cursor::new(Vec::new()));
+    for (name, bytes) in entries {
+        zw.start_file(*name, zip::write::SimpleFileOptions::default()).unwrap();
+        zw.write_all(bytes).unwrap();
+    }
+    zw.finish().unwrap().into_inner()
+}
+
+fn document_xml(bytes: &[u8]) -> String {
+    let mut z = zip::ZipArchive::new(std::io::Cursor::new(bytes)).unwrap();
+    let mut s = String::new();
+    z.by_name("word/document.xml").unwrap().read_to_string(&mut s).unwrap();
+    s
+}
+
+fn solid(rgb: &str) -> String {
+    format!(r#"<a:solidFill><a:srgbClr val="{rgb}"/></a:solidFill>"#)
+}
+
+/// A `wsp` at (`x`, `y`) EMUs in its group, `cx` × `cy`. `text` gives it a text box.
+fn shape(n: &Names, prst: &str, (x, y, cx, cy): (u32, u32, u32, u32), fill: &str, text: Option<&str>) -> String {
+    let s = n.shape;
+    let connector = prst.contains("Connector");
+    let nv = if connector {
+        format!("<{s}:cNvCnPr/>")
+    } else if text.is_some() {
+        format!(r#"<{s}:cNvSpPr txBox="1"/>"#)
+    } else {
+        format!("<{s}:cNvSpPr/>")
+    };
+    let fill = if connector { String::new() } else { solid(fill) };
+    let line =
+        if connector { r#"<a:ln w="25400"><a:solidFill><a:srgbClr val="117A65"/></a:solidFill><a:tailEnd type="triangle"/></a:ln>"# } else { "" };
+    let txbx = match text {
+        Some(t) => {
+            let c = n.content;
+            format!(r#"<{s}:txbx><{c}><w:p><w:r><w:t>{t}</w:t></w:r></w:p></{c}></{s}:txbx>"#)
+        }
+        None => String::new(),
+    };
+    format!(
+        r#"<{s}:wsp><{s}:cNvPr id="{cx}" name="{prst}"/>{nv}<{s}:spPr><a:xfrm><a:off x="{x}" y="{y}"/><a:ext cx="{cx}" cy="{cy}"/></a:xfrm><a:prstGeom prst="{prst}"><a:avLst/></a:prstGeom>{fill}{line}</{s}:spPr>{txbx}<{s}:bodyPr anchor="ctr"/></{s}:wsp>"#
+    )
+}
+
+/// Drawing `id`, anchored `x` EMUs from the page's left edge, 1 in × ½ in.
+fn anchor(id: u32, x: u32, uri: &str, payload: &str) -> String {
+    format!(
+        r#"<w:r><w:drawing><wp:anchor distT="0" distB="0" distL="0" distR="0" simplePos="0" relativeHeight="{id}" behindDoc="0" locked="0" layoutInCell="1" allowOverlap="1"><wp:simplePos x="0" y="0"/><wp:positionH relativeFrom="page"><wp:posOffset>{x}</wp:posOffset></wp:positionH><wp:positionV relativeFrom="page"><wp:posOffset>914400</wp:posOffset></wp:positionV><wp:extent cx="914400" cy="457200"/><wp:effectExtent l="0" t="0" r="0" b="0"/><wp:wrapNone/><wp:docPr id="{id}" name="Drawing {id}"/><wp:cNvGraphicFramePr/><a:graphic><a:graphicData uri="{uri}">{payload}</a:graphicData></a:graphic></wp:anchor></w:drawing></w:r>"#
+    )
+}
+
+/// A rectangle, a text box, an arrow, a group holding a nested group, and a canvas holding a
+/// shape and a group, each 1 in × ½ in.
+fn package(n: &Names) -> Vec<u8> {
+    const WPS: &str = "http://schemas.microsoft.com/office/word/2010/wordprocessingShape";
+    const WPG: &str = "http://schemas.microsoft.com/office/word/2010/wordprocessingGroup";
+    const WPC: &str = "http://schemas.microsoft.com/office/word/2010/wordprocessingCanvas";
+    let full = (0, 0, 914_400, 457_200);
+    let half = (0, 0, 457_200, 457_200);
+    let g = n.group;
+    let group_xfrm = |(x, y, cx, cy): (u32, u32, u32, u32)| {
+        format!(
+            r#"<{g}:cNvGrpSpPr/><{g}:grpSpPr><a:xfrm><a:off x="{x}" y="{y}"/><a:ext cx="{cx}" cy="{cy}"/><a:chOff x="0" y="0"/><a:chExt cx="{cx}" cy="{cy}"/></a:xfrm>{}</{g}:grpSpPr>"#,
+            solid("F39C12")
+        )
+    };
+    let group = format!(
+        "<{g}:wgp>{}{}<{g}:grpSp>{}{}</{g}:grpSp></{g}:wgp>",
+        group_xfrm(full),
+        shape(n, "ellipse", half, "28B463", None),
+        group_xfrm((457_200, 0, 457_200, 457_200)),
+        shape(n, "triangle", half, "8E44AD", None),
+    );
+    let c = n.canvas;
+    let canvas = format!(
+        r#"<{c}:wpc><{c}:bg>{}</{c}:bg><{c}:whole><a:ln w="12700">{}</a:ln></{c}:whole>{}<{g}:wgp>{}{}</{g}:wgp></{c}:wpc>"#,
+        solid("FDF2E9"),
+        solid("A04000"),
+        shape(n, "roundRect", (45_720, 45_720, 365_760, 365_760), "16A085", None),
+        group_xfrm((457_200, 45_720, 365_760, 365_760)),
+        shape(n, "diamond", (0, 0, 365_760, 365_760), "2E86C1", None),
+    );
+    let drawings = [
+        anchor(1, 914_400, WPS, &shape(n, "rect", full, "C0392B", None)),
+        anchor(2, 2_286_000, WPS, &shape(n, "rect", full, "EAF2F8", Some("Strict text box"))),
+        anchor(3, 3_657_600, WPS, &shape(n, "straightConnector1", full, "", None)),
+        anchor(4, 5_029_200, WPG, &group),
+        anchor(5, 6_400_800, WPC, &canvas),
+    ];
+    let doc = format!(
+        r#"<?xml version="1.0" encoding="UTF-8" standalone="yes"?><w:document {}><w:body><w:p><w:r><w:t>Shapes</w:t></w:r>{}</w:p><w:sectPr><w:pgSz w:w="12240" w:h="15840"/></w:sectPr></w:body></w:document>"#,
+        n.ns,
+        drawings.concat()
+    );
+    zip(&[("_rels/.rels", n.rels.as_bytes()), ("word/document.xml", doc.as_bytes())])
+}
+
+fn objects(d: &Document) -> Vec<InlineObject> {
+    d.body[0].as_para().unwrap().objects.clone()
+}
+
+/// The text of each text box story, in reading order.
+fn box_texts(d: &Document) -> Vec<String> {
+    let mut ids = Vec::new();
+    d.objects_in_reading_order(&mut |o| ids.extend(o.text_boxes()));
+    ids.iter()
+        .filter_map(|id| d.parts.get(id).filter(|p| p.kind == PartKind::TextBox))
+        .map(|p| p.blocks.iter().filter_map(|b| b.as_para()).map(|p| p.text.clone()).collect::<Vec<_>>().join("\n"))
+        .collect()
+}
+
+#[test]
+fn strict_shapes_read_like_their_transitional_twins() {
+    let transitional = wordcraft_docx::read(&package(&TRANSITIONAL)).unwrap();
+    let want = objects(&transitional);
+    // The twin itself reads every drawing: a rectangle, a text box, an arrow, a group whose
+    // nested group is flattened into it, and a canvas with its background.
+    assert_eq!(want.len(), 5, "{want:#?}");
+    assert!(matches!(&want[0], InlineObject::Shape { story: None, freeform: None, .. }), "{:?}", want[0]);
+    assert!(matches!(&want[1], InlineObject::Shape { story: Some(_), .. }), "{:?}", want[1]);
+    assert!(matches!(&want[3], InlineObject::Group { children, canvas: None, .. } if children.len() == 2), "{:?}", want[3]);
+    assert!(
+        matches!(&want[4], InlineObject::Group { children, canvas: Some(c), .. } if children.len() == 2 && c.fill.is_some() && c.stroke.is_some()),
+        "{:?}",
+        want[4]
+    );
+    assert_eq!(box_texts(&transitional), ["Strict text box"]);
+    for (name, names) in [("Word's Strict names", &STRICT_WORD), ("the Strict schema's names", &STRICT_SCHEMA)] {
+        let strict = wordcraft_docx::read(&package(names)).unwrap();
+        assert_eq!(objects(&strict), want, "{name}");
+        assert_eq!(box_texts(&strict), ["Strict text box"], "{name}");
+    }
+}
+
+#[test]
+fn strict_shapes_are_saved_and_reopened() {
+    let strict = wordcraft_docx::read(&package(&STRICT_WORD)).unwrap();
+    let saved = wordcraft_docx::write(&strict).unwrap();
+    let xml = document_xml(&saved);
+    for name in ["<wps:wsp>", "<wpg:wgp>", "<wpc:wpc>", "<wpc:bg>", "<w:txbxContent>", "<wps:cNvCnPr"] {
+        assert!(xml.contains(name), "{name} missing from the saved document");
+    }
+    let reopened = wordcraft_docx::read(&saved).unwrap();
+    assert_eq!(objects(&reopened), objects(&strict));
+    assert_eq!(box_texts(&reopened), ["Strict text box"]);
+}

crates/docx/tests/it/text_boxes.rs

@@ -132,6 +132,15 @@ fn vml_text_boxes_keep_their_inset_fit_and_anchor() {
         shape("", r#"o:insetmode="auto" inset="0,0,0,0""#),
         // In a group, a member's margins go with that member's text, not an empty box's before it.
         format!(r#"<w:p><w:r><w:pict><v:group style="width:100pt;height:40pt"><v:shape><v:textbox inset="9pt,9pt,9pt,9pt"/></v:shape><v:shape><v:textbox inset="0,0,0,0">{content}</v:textbox></v:shape></v:group></w:pict></w:r></w:p>"#),
+        // Text direction (#5193), as Word for Windows reads it: `vertical` and
+        // `vertical-ideographic` are `vert`; a `bottom-to-top` alt is `vert270` with any
+        // `layout-flow` or none; `horizontal-ideographic` stays horizontal.
+        shape("", r#"style="layout-flow:vertical""#),
+        shape("", r#"style="layout-flow:vertical;mso-layout-flow-alt:bottom-to-top;mso-next-textbox:#Text Box 4""#),
+        shape("", r#"style="layout-flow:vertical-ideographic""#),
+        shape("", r#"style="mso-layout-flow-alt:bottom-to-top""#),
+        shape("", r#"style="layout-flow:horizontal;mso-layout-flow-alt:bottom-to-top""#),
+        shape("", r#"style="layout-flow:horizontal-ideographic""#),
     ]
     .concat();
     let doc = format!(
@@ -140,13 +149,15 @@ fn vml_text_boxes_keep_their_inset_fit_and_anchor() {
     let rels = br#"<Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
     let d = wordcraft_docx::read(&zip(&[("_rels/.rels", rels), ("word/document.xml", doc.as_bytes())])).unwrap();
     let bodies: Vec<TextBody> = boxes(&d).iter().map(|id| d.parts[id].body).collect();
-    assert_eq!(bodies.len(), 6);
+    assert_eq!(bodies.len(), 12);
     assert_eq!(bodies[0].inset, Some([0, 0, 0, 0]));
     assert_eq!((bodies[1].inset, bodies[1].auto_fit, bodies[1].anchor), (Some([36_000, 45_720, 25_400, 45_720]), true, VAlign::Center));
     assert_eq!(bodies[2], TextBody::default());
     assert_eq!((bodies[3].inset, bodies[3].anchor), (Some([12_700, 45_720, 0, wordcraft_doc::MAX_TEXT_BOX_INSET]), VAlign::Bottom));
     assert_eq!(bodies[4].inset, None);
     assert_eq!(bodies[5].inset, Some([0, 0, 0, 0]));
+    let verts: Vec<TextVert> = bodies[6..].iter().map(|b| b.vert).collect();
+    assert_eq!(verts, [TextVert::Vert, TextVert::Vert270, TextVert::Vert, TextVert::Vert270, TextVert::Vert270, TextVert::Horz]);
     // Saved and read again, they stay.
     let back = wordcraft_docx::read(&wordcraft_docx::write(&d).unwrap()).unwrap();
     let again: Vec<TextBody> = boxes(&back).iter().map(|id| back.parts[id].body).collect();

crates/docx/tests/it/toc_ranges.rs

@@ -228,6 +228,7 @@ fn hyperlink_field_targets_follow_the_field_code() {
     for (para, link) in [
         (complex(r#"HYPERLINK "https://example.org/a" \l "sec""#), Some("https://example.org/a#sec")),
         (complex(r#"hyperlink \o "A tip" \t "_blank" "https://example.org/""#), Some("https://example.org/")),
+        (complex(r#"HYPERLINK "x \* Upper""#), Some("x * Upper")),
         (complex(r#"HYPERLINK "C:\\Docs\\My file.docx""#), Some(r"C:\Docs\My file.docx")),
         (simple(r#"HYPERLINK \l "Sec" \m"#), Some("#Sec")),
         (simple("HYPERLINK https://example.org/b"), Some("https://example.org/b")),

crates/docx/tests/it/twips_wrap.rs

@@ -0,0 +1,55 @@
+//! Paragraph indents and spacing written as plain twips beyond 16 bits: Word keeps only the low
+//! 16 bits (ECMA-376 Part 1 §17.3.1.12 `w:ind`, §17.3.1.33 `w:spacing`). Measured in Word: a
+//! `w:left` of 66256 starts the text 36 pt in, a `w:right` of 66976 leaves a 72 pt right indent.
+//! Fixtures are hand-written.
+
+use wordcraft_doc::Document;
+use wordcraft_doc::props::ParaProps;
+
+use crate::fixtures::zip;
+
+/// A package whose body is `body`.
+fn docx(body: &str) -> Vec<u8> {
+    let root = r#"<?xml version="1.0" encoding="UTF-8"?><Relationships xmlns="http://schemas.openxmlformats.org/package/2006/relationships"><Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="word/document.xml"/></Relationships>"#;
+    let doc = format!(
+        r#"<?xml version="1.0" encoding="UTF-8" standalone="yes"?><w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:body>{body}</w:body></w:document>"#
+    );
+    zip(&[("_rels/.rels", root.as_bytes()), ("word/document.xml", doc.as_bytes())])
+}
+
+fn props(body: &str) -> Vec<ParaProps> {
+    let d: Document = wordcraft_docx::read(&docx(body)).unwrap();
+    d.body.iter().filter_map(|b| b.as_para()).map(|p| p.props.clone()).collect()
+}
+
+#[test]
+fn indents_beyond_16_bits_wrap() {
+    let p = props(
+        r#"<w:p><w:pPr><w:ind w:left="66256"/></w:pPr><w:r><w:t>a</w:t></w:r></w:p>
+        <w:p><w:pPr><w:ind w:right="66976"/></w:pPr><w:r><w:t>b</w:t></w:r></w:p>
+        <w:p><w:pPr><w:ind w:start="65556" w:end="-65516"/></w:pPr><w:r><w:t>c</w:t></w:r></w:p>
+        <w:p><w:pPr><w:ind w:left="98304" w:hanging="65896"/></w:pPr><w:r><w:t>d</w:t></w:r></w:p>"#,
+    );
+    assert_eq!(p[0].indent_left, Some(36.0));
+    assert_eq!(p[1].indent_right, Some(72.0));
+    // 65556 -> 20 twips; -65516 -> 20 twips.
+    assert_eq!((p[2].indent_left, p[2].indent_right), (Some(1.0), Some(1.0)));
+    // 98304 = 0x18000 -> 0x8000 = -32768 as a signed 16-bit value; hanging 65896 -> 360.
+    assert_eq!((p[3].indent_left, p[3].indent_first), (Some(-1638.4), Some(-18.0)));
+}
+
+#[test]
+fn spacing_beyond_16_bits_wraps() {
+    let p = props(r#"<w:p><w:pPr><w:spacing w:before="65537" w:after="65776"/></w:pPr><w:r><w:t>a</w:t></w:r></w:p>"#);
+    assert_eq!((p[0].space_before, p[0].space_after), (Some(0.05), Some(12.0)));
+}
+
+#[test]
+fn in_range_and_unit_values_are_unchanged() {
+    let p = props(
+        r#"<w:p><w:pPr><w:ind w:left="32767" w:right="-32768"/><w:spacing w:before="65535"/></w:pPr><w:r><w:t>a</w:t></w:r></w:p>
+        <w:p><w:pPr><w:ind w:left="1in" w:right="720"/><w:spacing w:before="12pt"/></w:pPr><w:r><w:t>b</w:t></w:r></w:p>"#,
+    );
+    assert_eq!((p[0].indent_left, p[0].indent_right, p[0].space_before), (Some(1638.35), Some(-1638.4), Some(3276.75)));
+    assert_eq!((p[1].indent_left, p[1].indent_right, p[1].space_before), (Some(72.0), Some(36.0), Some(12.0)));
+}

crates/docx/tests/it/word_xml.rs

@@ -270,3 +270,12 @@ fn flat_opc_limits_count_what_is_not_kept() {
     // A part nothing reads, one element over what the XML tree reader takes in a part (8 000 000).
     assert!(limited(&format!(r#"<pkg:part pkg:name="/word/unused.xml"><pkg:xmlData><r>{}</r></pkg:xmlData></pkg:part>"#, "<y/>".repeat(8_000_000))));
 }
+
+/// A Word 2003 XML Document saved with Windows line ends: XML reads CR LF as one line feed (XML
+/// 1.0 §2.11), so a line break inside `w:t` is one space, not two.
+#[test]
+fn word_2003_xml_with_windows_line_ends_reads_one_space_per_line_break() {
+    let xml = "<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"yes\"?>\r\n<w:wordDocument xmlns:w=\"http://schemas.microsoft.com/office/word/2003/wordml\">\r\n<w:body><w:p><w:r><w:t>a\r\nb</w:t></w:r></w:p></w:body></w:wordDocument>\r\n";
+    let d = read_xml(xml.as_bytes()).unwrap();
+    assert_eq!(paras(&d)[0].plain_text(), "a b");
+}

crates/engine/src/cmd/caret.rs

@@ -323,6 +323,7 @@ fn select_all(s: &mut Session, v: &Value) -> CmdResult {
     }
     let st = s.sel.focus.story;
     s.sel = Selection { anchor: s.doc.start_of(st), focus: s.doc.end_of(st) };
+    s.pending = None;
     s.column = None;
     s.column_mode = false;
     sel_result(s)
@@ -332,6 +333,7 @@ fn select_range(s: &mut Session, v: &Value) -> CmdResult {
     let a = v.get("anchor").and_then(parse_pos).ok_or_else(|| CmdError::Params("bad `anchor`".into()))?;
     let b = v.get("focus").and_then(parse_pos).ok_or_else(|| CmdError::Params("bad `focus`".into()))?;
     s.sel = Selection { anchor: s.doc.clamp(&a), focus: s.doc.clamp(&b) };
+    s.pending = None;
     sel_result(s)
 }
 
@@ -392,6 +394,7 @@ fn select_unit(s: &mut Session, unit: Unit) -> CmdResult {
     if s.doc.para_at(&f).is_some() {
         let (a, b) = unit_range(s, &f, unit);
         s.sel = Selection { anchor: a, focus: b };
+        s.pending = None;
     }
     sel_result(s)
 }
@@ -486,6 +489,9 @@ fn select_text(s: &mut Session, v: &Value) -> CmdResult {
             if count == n {
                 s.sel =
                     Selection { anchor: Pos { story, path: path.clone(), off: i }, focus: Pos { story, path: path.clone(), off: i + needle.len() } };
+                // Formatting picked for the old caret doesn't carry over (as `select.word`): what
+                // replaces the selection takes its formatting.
+                s.pending = None;
                 return sel_result(s);
             }
         }

crates/engine/src/cmd/file.rs

@@ -253,27 +253,8 @@ fn open(s: &mut Session, v: &Value) -> CmdResult {
         );
     }
     let password = p::str(v, "password");
-    let (mut doc, encrypted) = if let Some(data) = p::str(v, "data") {
-        let bytes = super::insert::base64_decode(data).ok_or_else(|| CmdError::Params("bad base64".into()))?;
-        // An empty file (File Explorer's New › Microsoft Word Document makes one) opens as a new
-        // blank document that saves back to it (#1384).
-        if bytes.is_empty() {
-            (normal_document(s), false)
-        } else {
-            crate::io::open_bytes_in(path, &bytes, password, &s.region()).map_err(CmdError::Failed)?
-        }
-    } else if crate::io::is_empty_file(std::path::Path::new(path)) {
-        (normal_document(s), false)
-    } else {
-        crate::io::open_path_in(std::path::Path::new(path), password, &s.region()).map_err(CmdError::Failed)?
-    };
-    // Plain text and Markdown carry no language: they are in the editing language.
-    let ext = std::path::Path::new(path).extension().and_then(std::ffi::OsStr::to_str).unwrap_or("").to_ascii_lowercase();
-    if let Some(lang) = s.prefs.default_language.as_ref().or(s.editing_lang.as_ref())
-        && matches!(ext.as_str(), "txt" | "text" | "md" | "markdown")
-    {
-        Arc::make_mut(&mut doc.styles).default_chr.lang = Some(lang.clone());
-    }
+    // An empty file opens as a new blank document that saves back to it (#1384).
+    let (doc, encrypted) = read_document(s, path, v)?;
     match s.used_because() {
         None => s.set_document(doc),
         Some(why) => {
@@ -294,6 +275,53 @@ fn open(s: &mut Session, v: &Value) -> CmdResult {
     )
 }
 
+/// The document the file at `path` holds (or its bytes, base64 in `data`, with `path` only
+/// naming it), read with `password`, and whether it was encrypted: what `file.open` and
+/// `file.newFromTemplate` read. An empty file (File Explorer's New › Microsoft Word Document makes
+/// one) is a new blank document (#1384). Plain text and Markdown carry no language, so they are
+/// in the editing language.
+fn read_document(s: &Session, path: &str, v: &Value) -> Result<(Document, bool), CmdError> {
+    let password = p::str(v, "password");
+    let (mut doc, encrypted) = if let Some(data) = p::str(v, "data") {
+        let bytes = super::insert::base64_decode(data).ok_or_else(|| CmdError::Params("bad base64".into()))?;
+        if bytes.is_empty() {
+            (normal_document(s), false)
+        } else {
+            crate::io::open_bytes_in(path, &bytes, password, &s.region()).map_err(CmdError::Failed)?
+        }
+    } else if crate::io::is_empty_file(std::path::Path::new(path)) {
+        (normal_document(s), false)
+    } else {
+        crate::io::open_path_in(std::path::Path::new(path), password, &s.region()).map_err(CmdError::Failed)?
+    };
+    let ext = std::path::Path::new(path).extension().and_then(std::ffi::OsStr::to_str).unwrap_or("").to_ascii_lowercase();
+    if let Some(lang) = s.prefs.default_language.as_ref().or(s.editing_lang.as_ref())
+        && matches!(ext.as_str(), "txt" | "text" | "md" | "markdown")
+    {
+        Arc::make_mut(&mut doc.styles).default_chr.lang = Some(lang.clone());
+    }
+    Ok((doc, encrypted))
+}
+
+/// File › New from a template (#2852): a new, untitled document with the template's content,
+/// attached to the template as in Word, read as `file.open` reads a file (`data`, `password`).
+/// It opens beside the open documents, as `file.new` does. With `replaceBlank` an untouched blank
+/// document gives way to it, as to a file `file.open` opens: that is how a template the platform
+/// hands the app starts (#3433). The template is only read; the new document has no file, so
+/// saving it asks where to write, and a protected template's password doesn't come along.
+pub(crate) fn new_from_template(s: &mut Session, v: &Value) -> CmdResult {
+    let path = p::req_str(v, "path")?;
+    let (mut doc, _) = read_document(s, path, v)?;
+    doc.settings.attached_template = Some(path.to_string());
+    if p::bool(v, "replaceBlank") == Some(true) && s.used_because().is_none() {
+        s.set_document(doc);
+    } else {
+        s.add_document(doc);
+    }
+    s.dirty = true;
+    sel_result(s)
+}
+
 /// Close a document with no window to ask about it (agents, headless sessions).
 fn close_document(s: &mut Session, v: &Value) -> CmdResult {
     let id = match v.get("id").filter(|x| !x.is_null()) {

crates/engine/src/cmd/find_pattern.rs

@@ -198,7 +198,9 @@ fn special(query: &str, o: &Opts) -> (String, bool) {
         if (o.ignore_punct && is_punct(c)) || (o.ignore_space && is_space(c)) {
             continue;
         }
-        atoms.push(lit(c));
+        // A keyboard apostrophe also finds text produced by smart quotes. Explicit character
+        // codes above, and literal/regex/wildcard searches, keep exact character matching.
+        atoms.push(if c == '\'' { "['‘’]".into() } else { lit(c) });
     }
     let filler = match (o.ignore_punct, o.ignore_space) {
         (true, true) => r"[\p{P} \t\x{A0}]*",

crates/engine/src/cmd/insert/links.rs

@@ -63,6 +63,17 @@ pub(super) fn link(s: &mut Session, v: &Value) -> CmdResult {
         let end = s.doc.insert_text(&at, &text, &props)?;
         s.sel = Selection::caret(end);
         s.pending = Some(CharProps::default());
+    } else if let Some(text) = retyped_text(s, v) {
+        // Text to display changed: the selection is retyped with it, then linked (as in Word).
+        crate::cmd::paste_text(s, &text)?;
+        let end = s.sel.focus.clone();
+        let start = Pos { off: end.off.saturating_sub(text.len()), ..end.clone() };
+        s.doc.format_range(&start, &end, &|c| {
+            c.link = Some(url.clone());
+            c.style = style.clone();
+        })?;
+        s.sel = Selection::caret(end);
+        s.pending = Some(CharProps::default());
     } else {
         let (a, b) = s.sel.ordered();
         s.doc.format_range(&a, &b, &|c| {
@@ -73,6 +84,33 @@ pub(super) fn link(s: &mut Session, v: &Value) -> CmdResult {
     sel_result(s)
 }
 
+/// The `text` to retype a selection with before linking it: only when it differs from the selected
+/// text and both fit in one line of one paragraph without pictures or fields, outside any field's
+/// result. Otherwise (as Word shows `<<Selection in Document>>`) the selection keeps its text.
+fn retyped_text(s: &Session, v: &Value) -> Option<String> {
+    use wordcraft_doc::para::{COLUMN_BREAK, LINE_BREAK, OBJ, PAGE_BREAK};
+    /// What can't stand in one line of text: breaks, and pictures or fields.
+    const NOT_ONE_LINE: [char; 7] = [LINE_BREAK, '\r', '\u{2029}', '\u{b}', PAGE_BREAK, COLUMN_BREAK, OBJ];
+    let text = p::str(v, "text").filter(|t| !t.is_empty() && t.len() <= 4096)?;
+    if text.contains(NOT_ONE_LINE) {
+        return None;
+    }
+    let (a, b) = s.sel.ordered();
+    if a.story != b.story || a.path != b.path {
+        return None;
+    }
+    let shown = s.doc.para(a.story, &a.path)?.text.get(a.off..b.off)?;
+    if shown.contains(NOT_ONE_LINE) || shown == text {
+        return None;
+    }
+    // A selection in a field's result (between its markers, which it doesn't hold) is retyped by
+    // the next update of the field, not here.
+    if s.doc.field_ranges(a.story).iter().any(|r| r.start < b && a <= r.end) {
+        return None;
+    }
+    Some(text.to_string())
+}
+
 /// Whether `name` can name a bookmark: starts with a letter or underscore, contains only alphanumeric characters
 /// or underscores, and is at most 40 characters (#3491). Word allows leading underscores for hidden bookmarks (_Toc, _Ref).
 pub fn bookmark_name_ok(name: &str) -> bool {

crates/engine/src/cmd/lists.rs

@@ -452,6 +452,36 @@ mod tests {
         assert_eq!(numbered(&mut b), want);
     }
 
+    /// #5757: a list numbered in Thai digits shows them on the page, with the level's text around
+    /// them (`๑.`, `(๒)`), and keeps its format through a save to .docx and back.
+    #[test]
+    fn a_thai_numbered_list_shows_thai_digits_and_survives_docx() {
+        let mut s = Session::new(Document::new());
+        for (i, t) in ["หนึ่ง", "สอง", "สาม", "a", "b", "c", "d", "e", "f", "g", "h"].iter().enumerate() {
+            if i > 0 {
+                s.run("text.newParagraph", &json!({})).unwrap();
+            }
+            s.run("text.insert", &json!({"text": t})).unwrap();
+        }
+        s.run("select.all", &json!({})).unwrap();
+        let levels = json!([{"format": "thaiNumbers", "text": "%1.", "start": 1}]);
+        s.run("list.define", &json!({"levels": levels})).unwrap();
+        let page: Vec<Option<String>> = numbered(&mut s).into_iter().map(|x| x.0).collect();
+        let want: Vec<Option<String>> =
+            ["๑.", "๒.", "๓.", "๔.", "๕.", "๖.", "๗.", "๘.", "๙.", "๑๐.", "๑๑."].iter().map(|l| Some(l.to_string())).collect();
+        assert_eq!(page, want);
+        let back = wordcraft_docx::read(&wordcraft_docx::write(&s.doc).unwrap()).unwrap();
+        let mut b = Session::new(back);
+        let again: Vec<Option<String>> = numbered(&mut b).into_iter().map(|x| x.0).collect();
+        assert_eq!(again, want, "thaiNumbers survives .docx");
+        // In parentheses too.
+        let mut s = Session::new(Document::new());
+        s.run("text.insert", &json!({"text": "x"})).unwrap();
+        s.run("select.all", &json!({})).unwrap();
+        s.run("list.define", &json!({"levels": [{"format": "thaiNumbers", "text": "(%1)", "start": 2}]})).unwrap();
+        assert_eq!(numbered(&mut s)[0].0.as_deref(), Some("(๒)"));
+    }
+
     /// #1793: the other heading presets of Word's List Library, level by level.
     #[test]
     fn every_heading_preset_numbers_three_heading_levels() {
@@ -459,7 +489,8 @@ mod tests {
             ("headings", [Some("1"), Some("1.1"), Some("1.1.1")]),
             ("article", [Some("Article I."), Some("Section 1.01"), Some("(a)")]),
             ("outlineHeadings", [Some("I."), Some("A."), Some("1.")]),
-            ("chapter", [Some("Chapter 1"), None, None]),
+            // The lower levels are "none" levels with no text: an empty label, as in Word (#4985).
+            ("chapter", [Some("Chapter 1"), Some(""), Some("")]),
         ] {
             let mut s = headings("Heading3");
             s.run("list.multilevelPreset", &json!({"preset": preset})).unwrap();

crates/engine/src/cmd/objects/picture.rs

@@ -76,9 +76,11 @@ pub fn cropped_original_size(s: &Session, o: &InlineObject) -> Option<(f32, f32)
 pub(super) fn size(s: &mut Session, v: &Value) -> CmdResult {
     let (_, o) = selected(s).ok_or_else(|| CmdError::Disabled("no object".into()))?;
     let lock = p::bool(v, "lockAspect").unwrap_or(true);
-    let (mut w, mut h) = obj_size(&o);
+    let (w0, h0) = obj_size(&o);
+    let (mut w, mut h) = (w0, h0);
     let scale = p::f32(v, "scale");
-    if p::bool(v, "original").unwrap_or(false) {
+    let original = p::bool(v, "original").unwrap_or(false);
+    if original {
         let (ow, oh) =
             cropped_original_size(s, &o).ok_or_else(|| CmdError::Disabled("only a picture whose image can be read has an original size".into()))?;
         let k = scale.unwrap_or(100.0) / 100.0;
@@ -113,7 +115,16 @@ pub(super) fn size(s: &mut Session, v: &Value) -> CmdResult {
         (None, None) => true,
         _ => lock,
     };
-    let (w, h) = fit_size(w, h, keep);
+    // Sides set in proportion (from the original, or from a size whose sides both have a length)
+    // fit the limits together. Otherwise each side the call sets fits on its own, and a side it
+    // leaves as it was keeps its size, a straight line's 0 too.
+    let (w, h) = if keep && (original || (w0 > 0.0 && h0 > 0.0 && (w, h) != (w0, h0))) {
+        fit_size(w, h, true)
+    } else {
+        let (fw, fh) = fit_size(w, h, false);
+        let left = |new: f32, old: f32| !original && new == old;
+        (if left(w, w0) { w } else { fw }, if left(h, h0) { h } else { fh })
+    };
     // Any drawing: a group's members scale with its frame (its child space is kept); a chart or
     // SmartArt graphic lays itself out again at the new size.
     let theme = s.doc.settings.theme_colors.clone();

crates/engine/src/cmd/objects/specs.rs

@@ -13,7 +13,7 @@ pub fn specs() -> Vec<CommandSpec> {
     vec![
         CommandSpec::new("picture.size", "Size", "Picture Format › Size", size)
             .params(
-                r#"{"width"?: pt, "height"?: pt, "lockAspect"?: bool, "scale"?: percent, "original"?: bool} (scale: of the current size, or with original of the picture's original size, its pixels at 96 ppi less what is cropped; original alone restores that size; one side alone keeps the proportions unless lockAspect is false; 1–4000 pt: when the proportions are kept, a side past 4000 pt shrinks the other with it)"#,
+                r#"{"width"?: pt, "height"?: pt, "lockAspect"?: bool, "scale"?: percent, "original"?: bool} (scale: of the current size, or with original of the picture's original size, its pixels at 96 ppi less what is cropped; original alone restores that size; one side alone keeps the proportions unless lockAspect is false; 1–4000 pt: when the proportions are kept (from the original, or with both sides having a length), a side past 4000 pt shrinks the other with it; a side the call leaves as it was keeps its size, a straight line's 0 too)"#,
             )
             .when(has_sizable),
         CommandSpec::new("picture.crop", "Crop", "Picture Format › Size", crop)
@@ -132,14 +132,26 @@ pub fn specs() -> Vec<CommandSpec> {
             } else {
                 return Err(CmdError::Params("`path` or `data` required".into()));
             };
-            wordcraft_geom::picture::probe(&bytes).map_err(|e| match e {
-                wordcraft_geom::picture::Refusal::Unsupported(why) => CmdError::Failed(why.into()),
-                _ => CmdError::Failed("not a supported image".into()),
-            })?;
+            let (px_w, px_h) = wordcraft_geom::picture::probe(&bytes)
+                .map(|p| (p.width, p.height))
+                .map_err(|e| match e {
+                    wordcraft_geom::picture::Refusal::Unsupported(why) => CmdError::Failed(why.into()),
+                    _ => CmdError::Failed("not a supported image".into()),
+                })?;
             let key = s.doc.add_media(bytes.clone(), crate::cmd::insert::image_ext(&bytes));
             with_obj(s, |o| {
-                if let InlineObject::Image { media, .. } = o {
+                if let InlineObject::Image { media, w, h, crop, .. } = o {
                     *media = key.clone();
+                    // The replacement keeps the picture's width and takes its own aspect ratio, so a
+                    // picture of another shape isn't stretched into the old frame. The old crop
+                    // described the old picture, so it is dropped.
+                    if px_w > 0 && px_h > 0 && w.is_finite() && *w > 0.0 {
+                        let nh = *w * px_h as f32 / px_w as f32;
+                        if nh.is_finite() && nh > 0.0 {
+                            *h = nh;
+                        }
+                    }
+                    *crop = [0.0; 4];
                 }
             })
         })

crates/engine/src/cmd/objects/tests.rs

@@ -1034,6 +1034,66 @@ fn picture_size_past_the_limit_keeps_the_proportions() {
     }
 }
 
+/// A side the call leaves as it was keeps its size, even outside 1–4000 pt: a straight line's 0
+/// stays 0 when only its length is typed. Sides set in proportion (the original, or a size whose
+/// sides both have a length) fit the limits together; any other side the call sets, on its own.
+#[test]
+fn picture_size_keeps_the_side_it_leaves() {
+    let mut s = Session::new(wordcraft_doc::Document::new());
+    s.run("insert.shape", &json!({"kind": "straightConnector", "width": 100.0, "height": 0.0})).unwrap();
+    let size = |s: &Session| obj_size(&selected(s).unwrap().1);
+    // A side a scale sets on its own: 0 has no length to scale.
+    let scaled = |x: f32, k: f32| if x == 0.0 { 0.0 } else { (x * k).clamp(1.0, 4000.0) };
+    // Sizes with both sides follow each other: they fit 1–4000 pt by the one factor nearest 1, worked out by hand
+    // for the calls `width` locked, `height` locked, `scale` 50 and `scale` 200 with a width (in that order).
+    let states = [
+        ((100.0, 0.0), None),
+        ((0.0, 100.0), None),
+        ((5000.0, 0.0), None),
+        ((0.0, 5000.0), None),
+        ((100.0, 0.5), Some([(200.0, 1.0), (4000.0, 20.0), (200.0, 1.0), (200.0, 1.0)])),
+        ((5000.0, 50.0), Some([(200.0, 2.0), (4000.0, 40.0), (2500.0, 25.0), (200.0, 2.0)])),
+        ((100.0, 50.0), Some([(200.0, 100.0), (400.0, 200.0), (50.0, 25.0), (200.0, 100.0)])),
+    ];
+    for ((w0, h0), together) in states {
+        let follow = |i: usize, alone: (f32, f32)| together.map_or(alone, |t: [(f32, f32); 4]| t[i]);
+        let cases = [
+            (json!({"width": 200.0, "lockAspect": false}), (200.0, h0)),
+            (json!({"height": 200.0, "lockAspect": false}), (w0, 200.0)),
+            (json!({"width": 200.0}), follow(0, (200.0, h0))),
+            (json!({"height": 200.0}), follow(1, (w0, 200.0))),
+            (json!({"scale": 50}), follow(2, (scaled(w0, 0.5), scaled(h0, 0.5)))),
+            (json!({"scale": 200, "width": 200.0}), follow(3, (200.0, scaled(h0, 2.0)))),
+            (json!({"scale": 100}), (w0, h0)),
+            (json!({}), (w0, h0)),
+            // The Size dialog: both sides sent, the height as it was.
+            (json!({"width": 200.0, "height": h0, "lockAspect": false}), (200.0, h0)),
+            (json!({"width": 200.0, "height": 300.0}), (200.0, 300.0)),
+        ];
+        for (v, (ew, eh)) in cases {
+            with_obj(&mut s, |o| o.set_size(w0, h0)).unwrap();
+            s.run("picture.size", &v).unwrap();
+            let (w, h) = size(&s);
+            assert!((w - ew).abs() < 1e-3 && (h - eh).abs() < 1e-3, "{w0}×{h0} {v}: {w}×{h}, not {ew}×{eh}");
+        }
+    }
+    // Both sides of a picture's original size fit together, whatever size it had, even when the
+    // height comes out as it was: 4050×1012.5 pt is 4000×1000, not 4000×1012.5.
+    let data = super::super::insert::base64_encode(&png(400, 100));
+    s.run("insert.picture", &json!({"data": data})).unwrap();
+    for (w0, h0) in [(3000.0, 1012.5), (4050.0, 0.0), (100.0, 100.0)] {
+        with_obj(&mut s, |o| o.set_size(w0, h0)).unwrap();
+        s.run("picture.size", &json!({"original": true, "scale": 1350})).unwrap();
+        let (w, h) = size(&s);
+        assert!((w - 4000.0).abs() < 0.01 && (h - 1000.0).abs() < 0.01, "{w0}×{h0}: {w}×{h}");
+    }
+    // A height the original sets is set, though it is the one the picture had: 300×75 pt at
+    // 5400 % is 16200×4050, so with a width of 200 the height fits to 4000.
+    with_obj(&mut s, |o| o.set_size(100.0, 4050.0)).unwrap();
+    s.run("picture.size", &json!({"original": true, "scale": 5400, "width": 200.0, "lockAspect": false})).unwrap();
+    assert_eq!(size(&s), (200.0, 4000.0));
+}
+
 /// A crop handle drop: `picture.crop` with the new size is one edit, one undo step.
 #[test]
 fn picture_crop_with_size_is_one_undo_step() {

crates/engine/src/cmd/para.rs

@@ -183,9 +183,17 @@ pub fn specs() -> Vec<CommandSpec> {
             .key(if cfg!(target_os = "macos") { "Ctrl+Shift+Right" } else { "Alt+Shift+Right" }),
         CommandSpec::new("para.shading", "Shading", "Home › Paragraph", |s, v| {
             let c = p::str(v, "color").and_then(Rgb::parse);
+            // Text selected within one paragraph is shaded itself, as in Word, so No Color takes
+            // off a shading a file gave the text (#4685); a caret or paragraphs shade paragraphs.
+            // `paragraph` (Borders and Shading › Apply to: Paragraph) shades the paragraphs anyway.
+            if p::bool(v, "paragraph") != Some(true) && shades_text(s) {
+                return super::format::apply(s, &|x| x.shading = c);
+            }
             fmt(s, &|p| p.shading = c)
         })
-        .params(r#"{"color": "RRGGBB" | null}"#),
+        .params(
+            r#"{"color": "RRGGBB" | null, "paragraph"?: bool} (text selected inside one paragraph is shaded itself, as in Word; `paragraph: true` shades the paragraphs instead, as Borders and Shading › Apply to: Paragraph)"#,
+        ),
         CommandSpec::new("para.borders", "Borders", "Home › Paragraph", borders).params(
             r#"{"kind": "bottom|top|left|right|none|all|outside|inside|horizontalLine", "width"?: pt, "color"?: "RRGGBB", "style"?: "single|double|dotted|dashed|thick|triple|dotDash|wave", "sides"?: Sides} where Sides = {"top"|"left"|"bottom"|"right"|"between": {"style"?, "width"?: pt, "color"?: "RRGGBB"} | null}: exactly these sides (others off), instead of `kind`"#,
         ),
@@ -264,3 +272,21 @@ pub fn specs() -> Vec<CommandSpec> {
         }),
     ]
 }
+
+/// Whether Shading applies to the selected text rather than to paragraphs: a selection inside one
+/// paragraph, short of its mark. A selection reaching past the mark ends at the next paragraph's
+/// start, so all of a paragraph's text is still text, except in the last paragraph of its story,
+/// cell or box, where it stands for the paragraph with its mark. A selected picture, shape or text
+/// box is its object character, not text (a field's object character is).
+fn shades_text(s: &Session) -> bool {
+    let (a, b) = s.sel.ordered();
+    if a == b || a.story != b.story || a.path != b.path {
+        return false;
+    }
+    let Some(p) = s.doc.para_at(&a) else { return false };
+    if super::objects::object_selection(s).is_some() {
+        return false;
+    }
+    let last = s.doc.next_para(a.story, &a.path).is_none_or(|q| q.parent() != a.path.parent());
+    !(last && a.off == 0 && b.off >= p.len())
+}

crates/engine/src/cmd/references.rs

@@ -103,8 +103,9 @@ pub fn specs() -> Vec<CommandSpec> {
 
 /// A field instruction split into its name, positional arguments and switches (ECMA-376
 /// §17.16.1): `TOC \o "1-3" \h` → name `TOC`, switches `o` = `1-3` and `h`. Quoted text stays
-/// whole (without the quotes; `\"` and `\\` inside it are a quote and a backslash). A switch
-/// takes the word after it as its argument unless that word is another switch.
+/// whole (without the quotes; a backslash inside it makes the next character literal, as in Word:
+/// `\"` is a quote, `\\` a backslash, `\*` an asterisk). A switch takes the word after it as its
+/// argument unless that word is another switch.
 #[derive(Clone, Debug, Default, PartialEq)]
 pub struct FieldCode {
     pub name: String,
@@ -126,7 +127,7 @@ pub fn field_words(instr: &str) -> Vec<(String, bool)> {
             while let Some(c) = chars.next() {
                 match c {
                     '"' => break,
-                    '\\' if matches!(chars.peek(), Some('"' | '\\')) => w.extend(chars.next()),
+                    '\\' if chars.peek().is_some() => w.extend(chars.next()),
                     c => w.push(c),
                 }
             }

crates/engine/src/cmd/references/fields.rs

@@ -102,7 +102,9 @@ struct SeqCount {
 
 /// The fresh results of the unlocked SEQ fields in the body and in the text boxes anchored there
 /// ([`update_seq`]). A text box's fields count where the box is anchored, before the fields after
-/// the anchor in its paragraph and after those before it (measured in Word for Windows).
+/// the anchor in its paragraph and after those before it (measured in Word for Windows). A locked
+/// SEQ keeps its result but still counts, `\r` included; a SEQ in a tracked deletion keeps its
+/// result and doesn't count, with markup shown or not (measured in Word for Windows, exporting).
 fn seq_results(doc: &wordcraft_doc::Document) -> Vec<FieldResult> {
     let mut count = SeqCount::default();
     for path in doc.para_paths(StoryRef::Body) {
@@ -120,7 +122,11 @@ fn seq_results(doc: &wordcraft_doc::Document) -> Vec<FieldResult> {
 
 /// Number the SEQ fields of paragraph `p` (at `path` in `story`) and of the text boxes anchored in it.
 fn count_seq(doc: &wordcraft_doc::Document, story: StoryRef, path: &Path, p: &Paragraph, count: &mut SeqCount) {
-    for (k, o) in p.objects.iter().enumerate() {
+    // The runs come in order and the objects by offset, so one cursor walks the deleted ranges.
+    let mut deleted = p.deleted_ranges().into_iter().peekable();
+    for (k, (off, o)) in p.objects_with_offsets().enumerate() {
+        while deleted.next_if(|r| r.end <= off).is_some() {}
+        let in_deletion = deleted.peek().is_some_and(|r| r.contains(&off));
         for id in o.text_boxes() {
             if count.boxes.insert(id) {
                 let part = StoryRef::Part(id);
@@ -131,9 +137,9 @@ fn count_seq(doc: &wordcraft_doc::Document, story: StoryRef, path: &Path, p: &Pa
                 }
             }
         }
-        let InlineObject::Field { instr, locked: false, .. } = o else { continue };
+        let InlineObject::Field { instr, locked, .. } = o else { continue };
         let f = FieldCode::parse(instr);
-        if f.name != "SEQ" {
+        if f.name != "SEQ" || in_deletion {
             continue;
         }
         let Some(id) = f.args.first() else { continue };
@@ -149,6 +155,9 @@ fn count_seq(doc: &wordcraft_doc::Document, story: StoryRef, path: &Path, p: &Pa
         } else if !f.has('c') {
             n.0 = n.0.saturating_add(1);
         }
+        if *locked {
+            continue;
+        }
         let fmt = wordcraft_layout::fields::format_switch(instr).unwrap_or_default();
         let r = if f.has('h') { String::new() } else { wordcraft_layout::fields::field_number(instr, fmt, n.0) };
         count.out.push((story, path.clone(), k, r));

crates/engine/src/cmd/references/tests.rs

@@ -1,3 +1,4 @@
+use super::properties::{new_property_value, text_result};
 use super::*;
 
 fn fields(s: &Session) -> Vec<(String, String)> {
@@ -446,7 +447,7 @@ fn update_fields_apply_the_capitalization_switches() {
             "seventh",
             "Seventh",
             "BUDGET",
-            "new \\* Upper"
+            "new * Upper"
         ]
     );
 }
@@ -548,6 +549,85 @@ fn caption_for_an_object_in_a_cell_stays_in_the_cell() {
     assert_eq!(texts.get(1).map(String::as_str), Some("Figure 1"), "{texts:?}");
 }
 
+/// A fresh scratch directory for the file-name tests below.
+#[cfg(not(target_arch = "wasm32"))]
+fn filename_scratch_dir(tag: &str) -> std::path::PathBuf {
+    let nanos = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH).unwrap().as_nanos();
+    let dir = std::env::temp_dir().join(format!("wordcraft-filename-field-{tag}-{}-{nanos}", std::process::id()));
+    std::fs::create_dir_all(&dir).unwrap();
+    dir
+}
+
+/// #3796: after Save As to a new name, Update Field with the caret in a header shows the new
+/// name for FILENAME and the new full path for FILENAME \p.
+#[cfg(not(target_arch = "wasm32"))]
+#[test]
+fn filename_fields_in_a_header_follow_save_as() {
+    let dir = filename_scratch_dir("save-as");
+    let (mut s, hdr) = header_fields_session(wordcraft_doc::Document::from_text("body"), &["FILENAME", r"FILENAME \p"]);
+    let first = dir.join("report_v1.docx");
+    let second = dir.join("report_v2.docx");
+    s.run("file.save", &json!({"path": first.to_string_lossy()})).unwrap();
+    s.run("references.updateFields", &json!({})).unwrap();
+    let got = story_fields(&s, StoryRef::Part(hdr));
+    assert_eq!(got[0].1, "report_v1.docx");
+    assert_eq!(got[1].1, first.to_string_lossy());
+
+    s.run("file.saveAs", &json!({"path": second.to_string_lossy()})).unwrap();
+    s.run("references.updateFields", &json!({})).unwrap();
+    let got = story_fields(&s, StoryRef::Part(hdr));
+    assert_eq!(got[0].1, "report_v2.docx");
+    assert_eq!(got[1].1, second.to_string_lossy());
+    let _ = std::fs::remove_dir_all(&dir);
+}
+
+/// #3796: a Word-style `FILENAME \* MERGEFORMAT` field whose cached text names the old file
+/// shows the new name after Save As and Update Field.
+#[cfg(not(target_arch = "wasm32"))]
+#[test]
+fn filename_mergeformat_field_replaces_its_cached_name_after_save_as() {
+    let dir = filename_scratch_dir("mergeformat");
+    let (mut s, hdr) = header_fields_session(wordcraft_doc::Document::from_text("body"), &[r"FILENAME \* MERGEFORMAT"]);
+    let first = dir.join("report_v1.docx");
+    s.run("file.save", &json!({"path": first.to_string_lossy()})).unwrap();
+    // The cached text a file written earlier would carry.
+    let path = s.doc.para_paths(StoryRef::Part(hdr)).remove(0);
+    if let Some(InlineObject::Field { result, .. }) = s.doc.para_mut(StoryRef::Part(hdr), &path).unwrap().objects.first_mut() {
+        *result = "report_v1.docx".into();
+    }
+    s.run("file.saveAs", &json!({"path": dir.join("report_v2.docx").to_string_lossy()})).unwrap();
+    s.run("references.updateFields", &json!({})).unwrap();
+    assert_eq!(story_fields(&s, StoryRef::Part(hdr))[0].1, "report_v2.docx");
+    let _ = std::fs::remove_dir_all(&dir);
+}
+
+/// #3796: renaming the document (File > Info) changes what FILENAME shows on the next Update
+/// Field; the extension stays.
+#[cfg(not(target_arch = "wasm32"))]
+#[test]
+fn filename_field_in_a_header_follows_rename() {
+    let dir = filename_scratch_dir("rename");
+    let (mut s, hdr) = header_fields_session(wordcraft_doc::Document::from_text("body"), &["FILENAME"]);
+    s.run("file.save", &json!({"path": dir.join("report_v1.docx").to_string_lossy()})).unwrap();
+    s.run("file.rename", &json!({"name": "report_v3"})).unwrap();
+    s.run("references.updateFields", &json!({})).unwrap();
+    assert_eq!(story_fields(&s, StoryRef::Part(hdr))[0].1, "report_v3.docx");
+    let _ = std::fs::remove_dir_all(&dir);
+}
+
+/// A backslash in a quoted field argument makes the next character literal, so `\*` there is an
+/// asterisk and not the `\* Upper` switch (Word shows `x * Upper` for `TITLE "x \* Upper"`).
+#[test]
+fn quoted_field_argument_backslash_escapes() {
+    for (instr, shown) in [(r#"TITLE "x \* Upper""#, "x * Upper"), (r#"TITLE "x \\* Upper""#, r"x \* Upper"), (r#"TITLE "a \"b\" c""#, r#"a "b" c"#)]
+    {
+        let value = new_property_value(instr).unwrap();
+        assert_eq!(value, shown, "{instr}");
+        assert_eq!(text_result(instr, value), shown, "{instr}");
+    }
+    assert_eq!(text_result(r#"TITLE "ab" \* Upper"#, new_property_value(r#"TITLE "ab" \* Upper"#).unwrap()), "AB");
+}
+
 /// Output (PDF, print, page images) shows the body's unlocked SEQ, STYLEREF and REF fields with
 /// their fresh results, as Word fills them in when it prints or exports even with "Update fields
 /// before printing" off; NUMWORDS and a locked field keep their stored results there, and the
@@ -713,6 +793,45 @@ fn output_shows_references_to_bookmarks_outside_the_body() {
     assert_eq!(results, expected);
 }
 
+/// Printed or exported, a locked SEQ keeps its result but still counts, its `\r` too, and a SEQ
+/// in a tracked deletion keeps its result and doesn't count (measured in Word for Windows,
+/// exporting with markup shown and hidden: Del 1, 9, 2; Lock 1, 7, 3; R 7, 6).
+#[test]
+fn output_counts_locked_seq_fields_and_not_deleted_ones() {
+    let mut s = Session::new(wordcraft_doc::Document::from_text(""));
+    let fields_in = ["SEQ Del", "SEQ Del", "SEQ Del", "SEQ Lock", "SEQ Lock", "SEQ Lock", "SEQ R \\r 5", "SEQ R"];
+    for instr in fields_in {
+        s.run("insert.field", &json!({"instr": instr, "result": "9"})).unwrap();
+        s.run("text.insert", &json!({"text": " "})).unwrap();
+    }
+    let top = wordcraft_doc::Path::top(0);
+    let para = s.doc.para_mut(StoryRef::Body, &top).unwrap();
+    for (k, o) in para.objects.iter_mut().enumerate() {
+        if let InlineObject::Field { result, locked, .. } = o {
+            *locked = matches!(k, 4 | 6);
+            *result = if *locked { "7" } else { "9" }.into();
+        }
+    }
+    para.touch();
+    // The second Del field is a tracked deletion.
+    let off = s.doc.para(StoryRef::Body, &top).unwrap().object_offsets()[1];
+    let at = |off| Pos { story: StoryRef::Body, path: top.clone(), off };
+    s.doc.format_range(&at(off), &at(off + '\u{FFFC}'.len_utf8()), &|c| c.del = Some(1)).unwrap();
+    let out = output_fields(&s.doc).expect("the stale results change");
+    let shown: Vec<String> = out
+        .para(StoryRef::Body, &top)
+        .unwrap()
+        .objects
+        .iter()
+        .filter_map(|o| match o {
+            InlineObject::Field { result, .. } => Some(result.clone()),
+            _ => None,
+        })
+        .collect();
+    assert_eq!(shown, ["1", "9", "2", "1", "7", "3", "7", "6"]);
+    assert!(fields(&s).iter().all(|f| f.1 == "9" || f.1 == "7"), "the document keeps its results");
+}
+
 /// Printed or exported, a REF to a bookmark the document doesn't have shows Word's error
 /// (measured in Word for Windows: "Error! Reference source not found."); the editor and the
 /// document keep the stored result, and a REF showing "above" or "below" (`\p`) keeps it too.

crates/engine/src/cmd/review/changes.rs

@@ -54,11 +54,20 @@ fn resolve_para(
     // An inserted paragraph mark that is rejected, or a deleted one that is accepted, goes away:
     // the next paragraph joins this one (the split is undone). A paragraph left empty goes away
     // whole, so the next one keeps its own style and numbering, as in Word. The mark at the end
-    // of a story or cell can't go, so it just stops being a revision.
-    if ((del.is_some() && accept) || (ins.is_some() && !accept))
-        && (crate::cmd::remove_empty_para(s, story, path)? || crate::cmd::join_next_para(s, story, path)?)
-    {
-        return Ok(());
+    // of a story or cell can't go, so it just stops being a revision. An empty paragraph whose
+    // mark carries a section break goes whole too: its section joins the following one, as when
+    // the break is deleted (#5667). Joining instead gave the next paragraph its properties.
+    if (del.is_some() && accept) || (ins.is_some() && !accept) {
+        let sectioned = story == StoryRef::Body
+            && path.0.len() == 1
+            && s.doc.para(story, path).is_some_and(|p| p.is_empty() && p.section.is_some())
+            && crate::cmd::has_next_para(s, story, path);
+        if sectioned {
+            s.doc.remove_section_break(path.last());
+        }
+        if crate::cmd::remove_empty_para(s, story, path)? || crate::cmd::join_next_para(s, story, path)? {
+            return Ok(());
+        }
     }
     let para = s.doc.para_mut(story, path)?;
     para.mark.ins = None;

crates/engine/src/cmd/table.rs

@@ -195,7 +195,7 @@ pub fn specs() -> Vec<CommandSpec> {
         })
         .params(r#"{"text"?: string, "title"?: string} (empty clears)"#)),
         t(CommandSpec::new("table.sort", "Sort", "Table Layout › Data", sort_table)
-            .params(r#"{"column"?: n, "descending"?: bool, "header"?: bool}"#)),
+            .params(r#"{"column"?: non-negative JSON integer (0-based, default 0; must name a column in the table), "descending"?: bool, "header"?: bool}"#)),
         t(CommandSpec::new("table.toText", "Convert to Text", "Table Layout › Data", to_text)
             .params(r#"{"separator"?: "tab|comma|paragraph|other", "char"?: string (with other)} (the selected rows, or the whole table)"#)),
         t(CommandSpec::new("table.formula", "Formula", "Table Layout › Data", formula).params(r#"{"formula"?: "=SUM(ABOVE)"}"#)),

crates/engine/src/cmd/table/convert.rs

@@ -8,11 +8,18 @@ use super::{
 /// Sort the table's rows by a column. Numbers are read in each cell's language (`1,5` is one and
 /// a half in Polish) and sort first, numerically; other text sorts in the column's language.
 pub(super) fn sort_table(s: &mut Session, v: &Value) -> CmdResult {
-    let col = p::u64(v, "column").unwrap_or(0) as usize;
+    let column = match v.get("column") {
+        None => 0,
+        Some(value) => value.as_u64().ok_or_else(|| CmdError::Params("`column` must be a non-negative JSON integer".into()))?,
+    };
     let desc = p::bool(v, "descending").unwrap_or(false);
     let header = p::bool(v, "header");
     let (tp, _, _) = cell(s)?;
     let t = s.doc.table(s.sel.focus.story, &tp).ok_or_else(|| CmdError::Failed("no table".into()))?;
+    let col = usize::try_from(column)
+        .ok()
+        .filter(|&col| t.rows.iter().any(|row| col < row.cells.len()))
+        .ok_or_else(|| CmdError::Params("`column` is outside the table".into()))?;
     let skip = usize::from(header.unwrap_or(t.rows.first().is_some_and(|r| r.props.header))).min(t.rows.len());
     let column = || t.rows.iter().skip(skip).map(|r| r.cells.get(col));
     let lang = column().flatten().find_map(|c| cell_lang(&s.doc, c)).or_else(|| crate::cmd::default_lang(s));

crates/engine/src/cmd/tools/specs.rs

@@ -121,17 +121,11 @@ pub fn specs() -> Vec<CommandSpec> {
         CommandSpec::new("file.recover", "Recover Unsaved Documents", "File › Info", |s, v| s.recover(v))
         .params(r#"{"restore"?: recoveryId, "discard"?: recoveryId, "refresh"?: true}; desktop only; returns {enabled,busy,documents,error}; restore loads asynchronously without a file path"#)
         .pure(),
-        CommandSpec::new("file.newFromTemplate", "New from Template", "File › New", |s, v| {
-            let path = p::req_str(v, "path")?;
-            let mut doc = crate::io::open_path(std::path::Path::new(path)).map_err(CmdError::Failed)?;
-            // As in Word, the new document is attached to the template it came from (#2852).
-            doc.settings.attached_template = Some(path.to_string());
-            s.add_document(doc);
-            s.dirty = true;
-            sel_result(s)
-        })
-        .params(r#"{"path": string} → a new, untitled document from the template, attached to it"#)
-        .pure(),
+        CommandSpec::new("file.newFromTemplate", "New from Template", "File › New", crate::cmd::file::new_from_template)
+            .params(
+                r#"{"path": string, "data"?: base64 (the template's bytes; `path` then only names it), "password"?: string (for a password-protected template), "replaceBlank"?: bool (an untouched blank document gives way to it, as to a file file.open opens)} → a new, untitled document from the template, attached to it"#,
+            )
+            .pure(),
         CommandSpec::new("file.saveTemplate", "Save as Template", "File › Save As", |s, v| {
             let path = p::req_str(v, "path")?;
             // A .dotm keeps its name (#2852); anything else becomes a .dotx.

crates/engine/src/cmd/user_templates.rs

@@ -337,6 +337,73 @@ mod tests {
         let _ = std::fs::remove_dir_all(&d);
     }
 
+    /// #3433: a template the platform hands the app takes an untouched blank document's place
+    /// (`replaceBlank`), as a file `file.open` opens does; otherwise, and without the flag, the
+    /// new document opens beside the others, as File › New's does. Either way the template on
+    /// disk is only read: the new document has no path, so a save can't write over it.
+    #[test]
+    fn new_from_template_replaces_an_untouched_blank_document_only_when_asked() {
+        let d = scratch("replace-blank");
+        let path = d.join("Letterhead.dotx");
+        template(&path, "Letterhead text", "");
+        let bytes = std::fs::read(&path).unwrap();
+        let from = json!({"path": path.to_string_lossy(), "replaceBlank": true});
+
+        let mut s = Session::new(wordcraft_doc::Document::new());
+        s.run("file.newFromTemplate", &from).unwrap();
+        assert_eq!(s.documents().len(), 1, "the untouched blank document gave way");
+        assert_eq!(s.path, None);
+        assert!(s.dirty);
+        assert_eq!(s.doc.settings.attached_template.as_deref(), Some(path.to_string_lossy().as_ref()));
+        assert!(s.doc.plain_text(wordcraft_doc::StoryRef::Body).contains("Letterhead text"));
+        let saved = s.run("file.save", &json!({})).unwrap();
+        assert_eq!(saved["saved"], json!(false), "no file to save to: Save As asks where");
+        assert_eq!(std::fs::read(&path).unwrap(), bytes, "the template is never written over");
+
+        // The new document has unsaved changes, so a second one opens beside it.
+        s.run("file.newFromTemplate", &from).unwrap();
+        assert_eq!(s.documents().len(), 2);
+        // Without the flag, even an untouched blank document stays (File › New).
+        let mut s = Session::new(wordcraft_doc::Document::new());
+        s.run("file.newFromTemplate", &json!({"path": path.to_string_lossy()})).unwrap();
+        assert_eq!(s.documents().len(), 2);
+        assert_eq!(std::fs::read(&path).unwrap(), bytes);
+        let _ = std::fs::remove_dir_all(&d);
+    }
+
+    /// #3433: a protected template asks for its password like `file.open`, and the document made
+    /// from it isn't protected; a template's bytes can come in `data` (a file name that isn't
+    /// valid Unicode, #1440), its path then only naming it.
+    #[test]
+    #[cfg(not(target_arch = "wasm32"))]
+    fn new_from_template_reads_a_protected_template_and_bytes_like_file_open() {
+        let d = scratch("protected");
+        let path = d.join("Locked.dotx");
+        let mut t = Session::new(wordcraft_doc::Document::from_text("Locked template"));
+        t.run("file.save", &json!({"path": path.to_string_lossy(), "password": "sesame"})).unwrap();
+        let name = path.to_string_lossy();
+
+        let mut s = Session::new(wordcraft_doc::Document::new());
+        let e = s.run("file.newFromTemplate", &json!({"path": name})).unwrap_err().to_string();
+        assert!(crate::io::needs_password(&e), "{e}");
+        let e = s.run("file.newFromTemplate", &json!({"path": name, "password": "wrong"})).unwrap_err().to_string();
+        assert!(crate::io::wrong_password(&e), "{e}");
+        assert_eq!(s.documents().len(), 1, "nothing opened yet");
+        s.run("file.newFromTemplate", &json!({"path": name, "password": "sesame", "replaceBlank": true})).unwrap();
+        assert!(s.doc.plain_text(wordcraft_doc::StoryRef::Body).contains("Locked template"));
+        assert_eq!((s.path.as_ref(), s.password.is_some()), (None, false));
+
+        let plain = d.join("Plain.dotx");
+        template(&plain, "From bytes", "");
+        let data = crate::cmd::insert::base64_encode(&std::fs::read(&plain).unwrap());
+        let mut s = Session::new(wordcraft_doc::Document::new());
+        s.run("file.newFromTemplate", &json!({"path": "Pl\u{FFFD}in.dotx", "data": data, "replaceBlank": true})).unwrap();
+        assert!(s.doc.plain_text(wordcraft_doc::StoryRef::Body).contains("From bytes"));
+        assert_eq!(s.doc.settings.attached_template.as_deref(), Some("Pl\u{FFFD}in.dotx"));
+        assert_eq!(s.path, None);
+        let _ = std::fs::remove_dir_all(&d);
+    }
+
     #[test]
     fn the_templates_folder_option_is_saved_and_checked() {
         let d = scratch("option");

crates/engine/src/io.rs

@@ -771,8 +771,8 @@ pub fn list_dir(path: &std::path::Path, max: usize) -> Result<(Vec<DirEntry>, bo
 #[cfg(not(target_arch = "wasm32"))]
 pub fn package_title(path: &std::path::Path, max_file: u64, max_part: u64) -> Option<String> {
     crate::file_access::check_read(&path.to_string_lossy()).ok()?;
-    let file = std::fs::File::open(path).ok()?;
-    if file.metadata().ok()?.len() > max_file {
+    let (file, meta) = open_regular_file(path).ok()?;
+    if meta.len() > max_file {
         return None;
     }
     wordcraft_docx::package_title(std::io::BufReader::new(file), max_part)

crates/engine/src/session/documents.rs

@@ -390,7 +390,7 @@ impl Session {
         let at = self.others.iter().position(|d| d.document_id == id).ok_or_else(|| unknown(id))?;
         let mut d = self.others.swap_remove(at);
         self.swap_doc(&mut d);
-        self.view.read_page_color = self.prefs.read_page_color;
+        self.view.follow_prefs(&self.prefs);
         self.others.push(d);
         Ok(())
     }
@@ -471,7 +471,7 @@ impl Session {
         doc.ensure_nonempty();
         let mut view = self.view.clone();
         view.outline.collapsed.clear();
-        view.read_page_color = self.prefs.read_page_color;
+        view.follow_prefs(&self.prefs);
         let mut d = DocState::new(doc, untitled, view);
         let id = d.document_id;
         self.order.push(id);

crates/engine/src/session/snapshot.rs

@@ -32,7 +32,7 @@ impl Session {
         self.hf_return = None;
         self.go_back_step = 0;
         self.view.outline.collapsed.clear();
-        self.view.read_page_color = self.prefs.read_page_color;
+        self.view.follow_prefs(&self.prefs);
         // A copy the replaced document kept stays in the recovery list.
         #[cfg(not(target_arch = "wasm32"))]
         {

crates/engine/src/session/views.rs

@@ -135,7 +135,7 @@ impl Session {
         let current = document_id == self.document_id && rev == self.rev;
         self.sel = sel;
         self.view = view;
-        self.view.read_page_color = self.prefs.read_page_color;
+        self.view.follow_prefs(&self.prefs);
         self.pending = pending;
         self.goal_x = goal_x;
         self.page_hint = page_hint;

crates/engine/src/tests.rs

@@ -12,6 +12,7 @@ use crate::{Session, cmd};
 
 mod bidi;
 mod documents;
+mod find_apostrophes;
 mod font_sizes;
 mod fonts;
 mod ime_noop;
@@ -4681,10 +4682,8 @@ fn typed_paragraphs_are_capped_for_every_caller() {
     assert!(!body_texts(&s).iter().any(|p| p.contains('y')));
 }
 
-#[test]
-fn hostile_params_never_panic() {
-    let reg = cmd::registry();
-    let junk = [
+fn hostile_junk() -> [serde_json::Value; 8] {
+    [
         json!(null),
         json!({}),
         json!({"text": 5, "size": "x", "path": [], "pos": {"block": 9999, "off": 99999}}),
@@ -4693,11 +4692,23 @@ fn hostile_params_never_panic() {
         json!({"page": 0, "x": 1e30, "y": -1e30, "extend": 1}),
         json!({"width": 1e-9, "height": 1e30, "top": -1e9, "left": 1e9}),
         json!({"section": {"pageW": -1e308, "pageH": 0, "marginTop": 1e30, "header": -5, "columns": {"count": 1000000, "space": -1e300, "widths": [[0, 0]]}}}),
-    ];
+    ]
+}
+
+/// `file.*`, `insert.picture` and `insert.textFromFile`: `hostile_file_params_never_panic` runs
+/// them, held to a folder of its own, and `hostile_params_never_panic` every other command.
+fn reaches_files(id: &str) -> bool {
+    id.starts_with("file.") || id == "insert.picture" || id == "insert.textFromFile"
+}
+
+#[test]
+fn hostile_params_never_panic() {
+    let reg = cmd::registry();
+    let junk = hostile_junk();
     for spec in reg.all() {
-        // These reach outside the session (files). Read Aloud (`review.readAloud`, `readAloud.*`)
-        // is fuzzed too: under `cfg(test)` its backend is the silent `Hold`, asserted below.
-        if spec.id.starts_with("file.") || spec.id == "insert.picture" || spec.id == "insert.textFromFile" {
+        // Read Aloud (`review.readAloud`, `readAloud.*`) is fuzzed too: under `cfg(test)` its
+        // backend is the silent `Hold`, asserted below.
+        if reaches_files(spec.id) {
             continue;
         }
         for j in &junk {
@@ -4711,6 +4722,127 @@ fn hostile_params_never_panic() {
     }
 }
 
+/// `file.*`, `insert.picture` and `insert.textFromFile` take junk params and paths to files they
+/// can't use (#3719). They may read only `in` and write only `out`, so a save can't replace a file
+/// another run is meant to find broken, and nothing outside those folders is touched.
+#[cfg(not(target_arch = "wasm32"))]
+#[test]
+fn hostile_file_params_never_panic() {
+    /// What is directly in `dir`: each name with its bytes (`None` for a folder).
+    fn contents(dir: &std::path::Path) -> std::collections::BTreeMap<String, Option<Vec<u8>>> {
+        let entries = std::fs::read_dir(dir).unwrap().map(|e| e.unwrap().path());
+        entries.map(|p| (p.file_name().unwrap().to_string_lossy().into_owned(), std::fs::read(&p).ok())).collect()
+    }
+    /// The names whose bytes differ, so a failure lists them instead of printing every file.
+    fn changed(before: &std::collections::BTreeMap<String, Option<Vec<u8>>>, dir: &std::path::Path) -> Vec<String> {
+        let after = contents(dir);
+        let names: std::collections::BTreeSet<_> = before.keys().chain(after.keys()).collect();
+        names.into_iter().filter(|n| before.get(*n) != after.get(*n)).cloned().collect()
+    }
+    let reg = cmd::registry();
+    let base = scratch_dir("hostile-files");
+    let (input, output, outside) = (base.join("in"), base.join("out"), base.join("outside"));
+    for dir in [&outside, &input.join("folder.docx"), &output.join("folder.docx")] {
+        std::fs::create_dir_all(dir).unwrap();
+    }
+    let at = |dir: &std::path::Path, name: &str| dir.join(name).to_string_lossy().into_owned();
+
+    // Fixed bytes, so a failure comes back on the next run.
+    let noise: Vec<u8> = (0..512u32).map(|i| (i.wrapping_mul(2_654_435_761) >> 24) as u8).collect();
+    let picture_exts = ["png", "jpg", "gif", "bmp", "svg", "emf"];
+    let mut paths = vec![];
+    for ext in crate::io::OPEN_EXTS.iter().chain(&picture_exts) {
+        let name = format!("random.{ext}");
+        std::fs::write(input.join(&name), &noise).unwrap();
+        paths.push(at(&input, &name));
+    }
+    paths.extend(crate::io::SAVE_EXTS.iter().map(|ext| at(&output, &format!("new.{ext}"))));
+    std::fs::write(input.join("empty"), b"").unwrap();
+    std::fs::write(input.join("empty.docx"), b"").unwrap();
+    std::fs::write(input.join("good.txt"), "Good text").unwrap();
+    std::fs::write(input.join("good.png"), tiny_png()).unwrap();
+    paths.extend(["empty", "empty.docx", "folder.docx", "missing", "missing.docx"].map(|name| at(&input, name)));
+    paths.extend(["folder.docx", "missing/new.docx", &format!("{}.docx", "x".repeat(5000))].map(|name| at(&output, name)));
+    // Outside the roots, however it is spelled, and not paths at all.
+    paths.extend(["new.docx", "new"].map(|name| at(&outside, name)));
+    paths.push(format!("{}/../outside/new.docx", input.to_string_lossy()));
+    paths.extend(["hostile-relative.docx", "hostile-relative", "", "\0", "a\0b.docx"].map(String::from));
+
+    // Every path alone; each of these also with every other kind of junk param. The numbers that
+    // cost time (`copies`, `scale`) stay small: huge ones are #3720.
+    let mut params: Vec<serde_json::Value> = hostile_junk().into();
+    params.extend(paths.iter().map(|p| json!({"path": p})));
+    for name in [
+        "in/random.docx",
+        "in/empty.docx",
+        "in/folder.docx",
+        "in/missing.docx",
+        "in/good.txt",
+        "in/good.png",
+        "out/new.docx",
+        "out/new.pdf",
+        "out/new.png",
+    ] {
+        let path = at(&base, name);
+        params.extend([
+            json!({"path": path, "password": "x", "format": "wordml2003", "markup": "x", "confirmDamaged": 1}),
+            json!({"path": path, "password": 5, "format": 5, "markup": true}),
+            json!({"path": path, "page": 0, "scale": 0, "width": -1e308, "alt": 5}),
+            json!({"path": path, "data": "not base64!"}),
+            json!({"path": path, "data": "AAAA"}),
+            json!({"path": path, "pagesPerSheet": 7, "range": "x", "pages": "9-1,,", "parity": "x", "copies": 3, "paper": "x"}),
+            json!({"dir": path, "name": path, "template": path, "userTemplatesFolder": path, "id": 1e30, "restore": 1e30, "save": path}),
+        ]);
+    }
+
+    let before = contents(&input);
+    let working_dir = std::path::Path::new(".");
+    let working_dir_before = contents(working_dir);
+    let roots = crate::file_access::AutomationRoots::new(Some(&input), Some(&output)).unwrap();
+    let confined = |s: &mut Session, id: &str, v: &serde_json::Value| crate::file_access::confine(roots.as_ref(), || s.run(id, v));
+    for spec in reg.all() {
+        if !reaches_files(spec.id) {
+            continue;
+        }
+        for v in &params {
+            let mut s = Session::new(crate::sample::sample_document());
+            let r = confined(&mut s, spec.id, v);
+            assert!(!caught_panic(&reg, &r), "{} panicked on {v}: {r:?}", spec.id);
+            s.clamp_selection();
+            let _ = s.layout();
+        }
+    }
+
+    let nothing = Vec::<String>::new();
+    assert_eq!(changed(&before, &input), nothing, "the read root is only read");
+    assert_eq!(changed(&Default::default(), &outside), nothing, "nothing outside the write root");
+    // Every name counts: an empty path comes out as `.pdf` or `.dotx` here.
+    assert_eq!(changed(&working_dir_before, working_dir), nothing, "nothing in the working directory");
+
+    // The junk above met the files, not a closed door: usable files work through the same roots...
+    let mut s = s();
+    confined(&mut s, "file.open", &json!({"path": at(&input, "good.txt")})).unwrap();
+    assert_eq!(s.doc.plain_text(StoryRef::Body), "Good text");
+    confined(&mut s, "insert.textFromFile", &json!({"path": at(&input, "good.txt")})).unwrap();
+    confined(&mut s, "insert.picture", &json!({"path": at(&input, "good.png")})).unwrap();
+    let saved: [(&str, &str, &[u8]); 3] =
+        [("file.saveAs", "control.docx", b"PK\x03\x04"), ("file.exportPdf", "control.pdf", b"%PDF-"), ("file.exportPng", "control.png", b"\x89PNG")];
+    for (id, name, signature) in saved {
+        confined(&mut s, id, &json!({"path": at(&output, name)})).unwrap();
+        assert_eq!(std::fs::read(output.join(name)).unwrap().get(..signature.len()), Some(signature), "{name}");
+    }
+    // ...the noise is refused by the reader, and only a path outside by the roots.
+    let refused = |s: &mut Session, path: String| match confined(s, "file.open", &json!({"path": path})) {
+        Err(crate::CmdError::Failed(m)) => m,
+        r => panic!("not refused: {r:?}"),
+    };
+    let unreadable = refused(&mut s, at(&input, "random.docx"));
+    assert!(!unreadable.contains("automation path rejected"), "{unreadable}");
+    let rejected = refused(&mut s, at(&outside, "new.docx"));
+    assert!(rejected.contains("automation path rejected"), "{rejected}");
+    let _ = std::fs::remove_dir_all(&base);
+}
+
 /// A panic that `Session::run`'s guard turned into an error, in a command or one it ran: still a
 /// crash path (#3718). Only the guard's own `"{id} panicked: "` counts, not an error that echoes
 /// those words.
@@ -5615,6 +5747,39 @@ fn opening_a_named_pipe_or_device_is_refused_without_blocking() {
     assert!(device.unwrap_err().contains("not a file"));
 }
 
+#[cfg(unix)]
+#[test]
+fn reading_a_template_title_from_a_named_pipe_is_refused_without_blocking() {
+    let dir = scratch_dir("template-title-fifo");
+    let fifo = dir.join("pipe.dotx");
+
+    let made = std::process::Command::new("mkfifo").arg(&fifo).status();
+    if !made.is_ok_and(|s| s.success()) {
+        let _ = std::fs::remove_dir_all(&dir);
+        eprintln!("mkfifo unavailable; skipping");
+        return;
+    }
+
+    let (tx, rx) = std::sync::mpsc::channel();
+    let path = fifo.clone();
+
+    std::thread::spawn(move || {
+        let title = crate::io::package_title(&path, 64 << 20, 64 << 10);
+        let _ = tx.send(title);
+    });
+
+    let result = rx.recv_timeout(std::time::Duration::from_secs(10));
+
+    if result.is_err() {
+        // Unblock a stuck reader so the test thread can finish.
+        let _ = std::os::unix::fs::OpenOptionsExt::custom_flags(std::fs::OpenOptions::new().write(true), libc::O_NONBLOCK).open(&fifo);
+    }
+
+    let _ = std::fs::remove_dir_all(&dir);
+
+    assert_eq!(result.expect("reading a template title blocked on a named pipe"), None);
+}
+
 #[cfg(unix)]
 #[test]
 fn html_picture_that_is_a_named_pipe_is_skipped_without_blocking() {
@@ -7812,6 +7977,97 @@ fn colour_commands_take_any_custom_hex() {
     assert_eq!(s.doc.settings.page_color, wordcraft_doc::Rgb::parse("123ABC"));
 }
 
+/// #4685: Home › Paragraph › Shading on text selected inside a paragraph shades that text, as in
+/// Word, so No Color takes off a text shading a file gave it (an orange "highlight" from a .docx is
+/// `w:shd`, which Text Highlight Color › No Color leaves, in Word too). A caret or a whole
+/// paragraph still shades the paragraph.
+#[test]
+fn shading_selected_text_shades_the_text_and_no_color_clears_it() {
+    let orange = wordcraft_doc::Rgb::parse("FFA500");
+    let mut s = s();
+    run(&mut s, "text.insert", json!({"text": "Orange text here"}));
+    run(&mut s, "select.text", json!({"text": "Orange"}));
+    run(&mut s, "para.shading", json!({"color": "FFA500"}));
+    let p = s.doc.para_at(&s.sel.focus).unwrap();
+    assert_eq!((p.props.shading, p.props_of_char(0).shading, p.props_of_char(7).shading), (None, orange, None));
+    // Highlight › No Color doesn't take it off; Shading › No Color does.
+    run(&mut s, "format.highlight", json!({"color": "none"}));
+    assert_eq!(s.doc.para_at(&s.sel.focus).unwrap().props_of_char(0).shading, orange);
+    run(&mut s, "para.shading", json!({"color": null}));
+    assert_eq!(s.doc.para_at(&s.sel.focus).unwrap().props_of_char(0).shading, None);
+    // A caret, or the whole paragraph, shades the paragraph and leaves the text alone.
+    for whole in [false, true] {
+        let mut s = self::s();
+        run(&mut s, "text.insert", json!({"text": "Orange text here"}));
+        run(&mut s, if whole { "select.all" } else { "caret.left" }, json!({}));
+        run(&mut s, "para.shading", json!({"color": "FFA500"}));
+        let p = s.doc.para_at(&s.sel.focus).unwrap();
+        assert_eq!((p.props.shading, p.props_of_char(0).shading), (orange, None), "whole: {whole}");
+    }
+    // All the text of a paragraph that isn't the last is still text: a selection with its mark
+    // would end at the next paragraph's start.
+    let mut s = self::s();
+    run(
+        &mut s,
+        "document.setText",
+        json!({"text": "Word
+More"}),
+    );
+    run(&mut s, "select.text", json!({"text": "Word"}));
+    run(&mut s, "para.shading", json!({"color": "FFA500"}));
+    let p = s.doc.para_at(&s.sel.focus).unwrap();
+    assert_eq!((p.props.shading, p.props_of_char(0).shading), (None, orange));
+    // A selected text box shades its anchor's paragraph, not every line of the box.
+    let mut s = self::s();
+    run(&mut s, "text.insert", json!({"text": "Anchor"}));
+    run(&mut s, "insert.textBox", json!({"text": "Inside"}));
+    let id = s.doc.parts.iter().find(|(_, p)| p.kind == wordcraft_doc::PartKind::TextBox).map(|(id, _)| *id).unwrap();
+    let end = s.doc.text_box_anchor(id).unwrap();
+    let start = wordcraft_doc::Pos { off: end.off - wordcraft_doc::para::OBJ.len_utf8(), ..end.clone() };
+    run(&mut s, "select.range", json!({"anchor": start, "focus": end}));
+    run(&mut s, "para.shading", json!({"color": "FFA500"}));
+    let boxed = s.doc.parts.get(&id).and_then(|p| p.blocks.first()).and_then(|b| b.as_para()).unwrap();
+    assert_eq!(boxed.props_of_char(0).shading, None, "the box's text isn't shaded");
+    assert_eq!(s.doc.para_at(&end).unwrap().props.shading, orange);
+    // A selected field is text (its object character isn't a drawing): No Color clears its shading.
+    let mut s = self::s();
+    run(&mut s, "text.insert", json!({"text": "Page "}));
+    run(&mut s, "insert.pageNumber", json!({"position": "current"}));
+    let end = s.sel.focus.clone();
+    let start = wordcraft_doc::Pos { off: end.off - wordcraft_doc::para::OBJ.len_utf8(), ..end.clone() };
+    run(&mut s, "select.range", json!({"anchor": start, "focus": end}));
+    run(&mut s, "para.shading", json!({"color": "FFA500"}));
+    let p = s.doc.para_at(&end).unwrap();
+    assert_eq!((p.props.shading, p.props_of_char(start.off).shading), (None, orange), "the field is shaded");
+    run(&mut s, "para.shading", json!({"color": null}));
+    assert_eq!(s.doc.para_at(&end).unwrap().props_of_char(start.off).shading, None);
+    // Borders and Shading › Apply to: Paragraph shades the paragraph, whatever is selected.
+    run(&mut s, "select.text", json!({"text": "Page"}));
+    run(&mut s, "para.shading", json!({"color": "FFA500", "paragraph": true}));
+    let p = s.doc.para_at(&s.sel.focus).unwrap();
+    assert_eq!((p.props.shading, p.props_of_char(0).shading), (orange, None));
+}
+
+/// #4685 review: in a cell's last paragraph all of its text stands for the paragraph with its
+/// mark (a selection can't reach the next paragraph's start there), so Shading shades the
+/// paragraph; part of the text is text.
+#[test]
+fn shading_all_of_a_cells_last_paragraph_shades_the_paragraph() {
+    let orange = wordcraft_doc::Rgb::parse("FFA500");
+    for (from, to, paragraph) in [(0, 9, true), (0, 4, false)] {
+        let mut s = s();
+        run(&mut s, "insert.table", json!({"rows": 1, "cols": 1}));
+        run(&mut s, "text.insert", json!({"text": "Cell text"}));
+        let cell = s.sel.focus.clone();
+        assert!(cell.path.cell().is_some());
+        run(&mut s, "select.range", json!({"anchor": Pos { off: from, ..cell.clone() }, "focus": Pos { off: to, ..cell.clone() }}));
+        run(&mut s, "para.shading", json!({"color": "FFA500"}));
+        let p = s.doc.para_at(&cell).unwrap();
+        let want = if paragraph { (orange, None) } else { (None, orange) };
+        assert_eq!((p.props.shading, p.props_of_char(0).shading), want, "{from}..{to}");
+    }
+}
+
 /// #2738: Home › Underline › Underline Color sets the underline's own colour and leaves its style
 /// alone; `auto` drops it again. One undo step each, kept through a DOCX round trip, and junk is
 /// a params error that changes nothing.
@@ -10167,6 +10423,73 @@ fn column_texts(s: &Session) -> Vec<String> {
     t.rows.iter().map(|r| r.cells[0].blocks.iter().filter_map(|b| b.as_para()).map(|p| p.plain_text()).collect::<String>()).collect()
 }
 
+#[test]
+fn table_sort_rejects_invalid_columns_without_changing_rows_or_history() {
+    for column in [
+        json!(-1),
+        json!(0.5),
+        json!(0.0),
+        json!("0"),
+        json!(null),
+        json!(true),
+        json!([]),
+        json!({}),
+        json!(1),
+        json!(u32::MAX as u64 + 1),
+        json!(u64::MAX),
+    ] {
+        let mut s = column_table("en-US", &["Charlie", "Alpha", "Bravo"]);
+        s.close_typing();
+        run(&mut s, "text.insert", json!({"text": "!"}));
+        let before = s.doc.clone();
+        let selection = s.sel.clone();
+        let dirty = s.dirty;
+        assert!(s.run("table.sort", &json!({"column": column, "header": false})).is_err(), "column {column}");
+        assert_eq!(s.doc, before, "column {column}");
+        assert_eq!(s.sel, selection, "column {column}");
+        assert_eq!(s.dirty, dirty, "column {column}");
+        run(&mut s, "edit.undo", json!({}));
+        assert_eq!(column_texts(&s), ["Charlie", "Alpha", "Bravo"], "column {column}: refusal must not consume Undo");
+        run(&mut s, "edit.redo", json!({}));
+        assert_eq!(s.doc, before, "column {column}: Redo must restore the preceding edit");
+    }
+}
+
+#[test]
+fn table_sort_keeps_default_and_valid_columns() {
+    let mut s = s();
+    run(&mut s, "insert.table", json!({"rows": 3, "cols": 2}));
+    for (k, value) in ["Charlie", "1", "Alpha", "3", "Bravo", "2"].into_iter().enumerate() {
+        if k > 0 {
+            run(&mut s, "text.tab", json!({}));
+        }
+        run(&mut s, "text.insert", json!({"text": value}));
+    }
+    let original = s.doc.clone();
+    for params in [json!({"header": false}), json!({"column": 0, "header": false})] {
+        run(&mut s, "table.sort", params);
+        assert_eq!(column_texts(&s), ["Alpha", "Bravo", "Charlie"]);
+        run(&mut s, "edit.undo", json!({}));
+        assert_eq!(s.doc, original);
+    }
+    run(&mut s, "table.sort", json!({"column": 1, "header": false}));
+    assert_eq!(column_texts(&s), ["Charlie", "Bravo", "Alpha"]);
+    run(&mut s, "table.sort", json!({"column": 1, "header": false, "descending": true}));
+    assert_eq!(column_texts(&s), ["Alpha", "Bravo", "Charlie"]);
+    // A column may be missing in some rows. Its present cells still sort normally.
+    let at = s.doc.body.iter().position(|b| b.as_table().is_some()).unwrap() as u32;
+    let table = s.doc.table_mut(StoryRef::Body, &wordcraft_doc::Path(vec![at])).unwrap();
+    table.rows[0].cells.remove(1);
+    run(&mut s, "table.sort", json!({"column": 1, "header": false}));
+    assert_eq!(column_texts(&s), ["Charlie", "Bravo", "Alpha"]);
+
+    let mut s = column_table("en-US", &["Title", "3", "1", "2"]);
+    let at = s.doc.body.iter().position(|b| b.as_table().is_some()).unwrap() as u32;
+    s.doc.table_mut(StoryRef::Body, &wordcraft_doc::Path(vec![at])).unwrap().rows[0].props.header = true;
+    run(&mut s, "table.sort", json!({"column": 0}));
+    assert_eq!(column_texts(&s), ["Title", "1", "2", "3"]);
+}
+
 /// #836: table sort reads numbers in the text's language: `1,5` is one and a half in Polish.
 #[test]
 fn table_sort_reads_decimal_comma_in_polish() {
@@ -10494,6 +10817,44 @@ fn picture_change_keeps_the_replacement_format() {
     let _ = std::fs::remove_dir_all(&dir);
 }
 
+/// Change Picture keeps the width and takes the replacement's aspect ratio: swapping a figure for
+/// a revised one of a different shape must not stretch it into the old frame.
+#[cfg(not(target_arch = "wasm32"))]
+#[test]
+fn picture_change_keeps_width_and_takes_new_aspect() {
+    let dir = scratch_dir("picture-change-aspect");
+    let mut wide = Vec::new();
+    image::RgbImage::from_pixel(40, 20, image::Rgb([200, 30, 30])).write_to(&mut std::io::Cursor::new(&mut wide), image::ImageFormat::Png).unwrap();
+    let mut tall = Vec::new();
+    image::RgbImage::from_pixel(30, 60, image::Rgb([30, 30, 200])).write_to(&mut std::io::Cursor::new(&mut tall), image::ImageFormat::Png).unwrap();
+    let (first, second) = (dir.join("wide.png"), dir.join("tall.png"));
+    std::fs::write(&first, &wide).unwrap();
+    std::fs::write(&second, &tall).unwrap();
+
+    let mut s = Session::new(wordcraft_doc::Document::new());
+    run(&mut s, "insert.picture", json!({"path": first.to_string_lossy()}));
+    let size = |s: &Session| {
+        let mut out = (0.0f32, 0.0f32, [1.0f32; 4]);
+        s.doc.objects_in_reading_order(&mut |o| {
+            if let wordcraft_doc::para::InlineObject::Image { w, h, crop, .. } = o {
+                out = (*w, *h, *crop);
+            }
+        });
+        out
+    };
+    let (w0, h0, _) = size(&s);
+    assert!((w0 / h0 - 2.0).abs() < 1e-3, "inserted at its own 2:1 shape: {w0} x {h0}");
+    run(&mut s, "picture.crop", json!({"left": 0.1}));
+    let (w_cropped, _, crop) = size(&s);
+    assert!(crop[0] > 0.0, "cropped before the change");
+    run(&mut s, "picture.change", json!({"path": second.to_string_lossy()}));
+    let (w1, h1, crop) = size(&s);
+    assert!((w1 - w_cropped).abs() < 1e-3, "the width is kept: {w_cropped} -> {w1}");
+    assert!((h1 / w1 - 2.0).abs() < 1e-3, "the height follows the new 1:2 picture: {w1} x {h1}");
+    assert_eq!(crop, [0.0; 4], "the old picture's crop is dropped");
+    let _ = std::fs::remove_dir_all(&dir);
+}
+
 #[test]
 fn json_open_repairs_hostile_run_lengths() {
     use wordcraft_doc::{Block, Document};
@@ -12497,3 +12858,48 @@ fn page_down_extends_a_selection_that_ends_in_collapsed_outline_text() {
     assert!(s.sel.focus.path.0[0] > 6, "Page Down extends the selection a page on: {:?}", s.sel.focus);
     assert!(s.layout().caret(&s.sel.focus).is_some(), "to a place the outline shows");
 }
+
+/// Issue #5667's document: `Intro`, then an inserted `Appendix` paragraph whose mark carries a
+/// section break (Normal, indented), then `Body` in Heading 1. The whole of `Appendix` is
+/// tracked as an insertion by `rev`, or as a deletion when `delete`.
+fn sectioned_revision(delete: bool) -> Session {
+    use wordcraft_doc::{Revision, RevisionKind};
+    let mut s = Session::new(wordcraft_doc::Document::from_text("Intro\nAppendix\nBody"));
+    run(&mut s, "caret.set", json!({"pos": {"block": 2, "off": 0}}));
+    run(&mut s, "para.style", json!({"style": "Heading1"}));
+    let kind = if delete { RevisionKind::Delete } else { RevisionKind::Insert };
+    s.doc.revisions.push(Revision { kind, author: "Ada".into(), date: String::new(), move_name: None });
+    s.doc.last_section.margin_left = 72.0;
+    let mut sect = s.doc.last_section.clone();
+    sect.margin_left = 144.0;
+    let p = s.doc.para_mut(StoryRef::Body, &wordcraft_doc::Path(vec![1])).unwrap();
+    let len = p.len();
+    p.format(0, len, &|c| if delete { c.del = Some(0) } else { c.ins = Some(0) }).unwrap();
+    if delete {
+        p.mark.del = Some(0)
+    } else {
+        p.mark.ins = Some(0)
+    }
+    p.props.indent_left = Some(36.0);
+    p.section = Some(Box::new(sect));
+    s
+}
+
+#[test]
+fn resolving_away_an_empty_paragraph_with_a_section_break_keeps_the_next_paragraphs_properties() {
+    // #5667: the resolved paragraph was left empty but kept for its section break, so the next
+    // paragraph joined it and took its properties (Normal, its indent). It goes whole instead,
+    // as an empty paragraph without a break does, and its section joins the following one.
+    for (delete, cmd) in [(false, "review.rejectAll"), (true, "review.acceptAll")] {
+        let mut s = sectioned_revision(delete);
+        assert_eq!(s.doc.sections().len(), 2);
+        run(&mut s, cmd, json!({}));
+        assert_eq!(body_texts(&s), ["Intro", "Body"], "{cmd}");
+        let body = s.doc.para(StoryRef::Body, &wordcraft_doc::Path(vec![1])).unwrap();
+        assert_eq!(body.props.style.as_deref(), Some("Heading1"), "{cmd}: Body keeps its style");
+        assert_eq!(body.props.indent_left, None, "{cmd}: and not the removed paragraph's indent");
+        assert!(s.doc.body.iter().filter_map(|b| b.as_para()).all(|p| p.section.is_none()), "{cmd}: the break went with it");
+        assert_eq!(s.doc.sections().len(), 1, "{cmd}");
+        assert_eq!(s.doc.last_section.margin_left, 72.0, "{cmd}: the following section's setup stays");
+    }
+}

crates/engine/src/tests/documents.rs

@@ -346,3 +346,51 @@ fn input_method_joins_stay_with_their_window() {
     assert_eq!(text(&s).trim(), "Hello", "B's typing is its own undo step");
     drop(a);
 }
+
+/// Check spelling as you type is one setting for the app, not one per open document (#5740):
+/// turned off (File › Options › Proofing, or Review's command) with one document active, it is off
+/// in the others too, and switching to another document doesn't turn it back on, so what is saved
+/// for the next run is what was chosen, whichever document is active then.
+#[test]
+fn check_spelling_is_one_setting_for_every_open_document() {
+    let mut s = s();
+    let a = s.document_id();
+    run(&mut s, "file.new", json!({}));
+    let b = s.document_id();
+    run(&mut s, "file.new", json!({}));
+    let c = s.document_id();
+    assert!(s.view.proofing);
+    s.activate(a).unwrap();
+    run(&mut s, "file.options", json!({"checkSpelling": false}));
+    for id in [b, c, a] {
+        s.activate(id).unwrap();
+        assert!(!s.view.proofing, "off in document {id}");
+        assert!(!s.prefs.check_spelling);
+    }
+    run(&mut s, "review.proofing", json!({"value": true}));
+    for id in [b, c] {
+        s.activate(id).unwrap();
+        assert!(s.view.proofing, "back on in document {id}");
+    }
+    // A document opened now starts with the setting as it is.
+    run(&mut s, "review.proofing", json!({"value": false}));
+    run(&mut s, "file.new", json!({}));
+    assert!(!s.view.proofing);
+}
+
+/// A window's view put back after Check spelling as you type changed in another window takes the
+/// setting as it is now, not as it was when the window's view was taken (#5740): every view made
+/// active follows the saved preference.
+#[test]
+fn a_window_view_put_back_follows_the_spelling_setting() {
+    let mut s = s();
+    assert!(s.view.proofing);
+    let held = s.take_view();
+    run(&mut s, "file.options", json!({"checkSpelling": false}));
+    s.put_view(held);
+    assert!(!s.view.proofing, "the view taken with spelling on comes back with it off");
+    let held = s.take_view();
+    run(&mut s, "review.proofing", json!({"value": true}));
+    s.put_view(held);
+    assert!(s.view.proofing);
+}

crates/engine/src/tests/find_apostrophes.rs

@@ -0,0 +1,48 @@
+use super::*;
+
+/// Keyboard smart quotes stay searchable, and replacement keeps their original spelling.
+#[test]
+fn keyboard_apostrophe_finds_smart_quoted_typing() {
+    let mut s = s();
+    for ch in "Tau'va".chars() {
+        run(&mut s, "text.insert", json!({"text": ch.to_string()}));
+    }
+    assert_eq!(text(&s), "Tau’va");
+    for query in ["Tau'", "Tau'v", "Tau'va", "tau'VA"] {
+        assert_eq!(run(&mut s, "edit.find", json!({"text": query}))["count"], 1, "{query}");
+    }
+    assert_eq!(run(&mut s, "edit.find", json!({"text": "tau'va", "matchCase": true}))["count"], 0);
+    assert_eq!(run(&mut s, "edit.replaceAll", json!({"text": "Tau'va", "with": "[^&]", "matchCase": false}))["replaced"], 1);
+    assert_eq!(text(&s), "[Tau’va]");
+    run(&mut s, "edit.undo", json!({}));
+    assert_eq!(text(&s), "Tau’va");
+    run(&mut s, "edit.redo", json!({}));
+    assert_eq!(text(&s), "[Tau’va]");
+}
+
+/// Normal search uses raw UTF-8 ranges while explicit syntax and modifier letters stay exact.
+#[test]
+fn normal_apostrophe_search_keeps_exact_modes_and_original_byte_ranges() {
+    let mut s = s();
+    let original = "🙂 Tau'va Tau’va Tau‘va Tauʼva";
+    run(&mut s, "document.setText", json!({"text": original}));
+    let found = run(&mut s, "edit.find", json!({"text": "Tau'va", "wholeWord": true}));
+    assert_eq!(found["count"], 3);
+    let ranges: Vec<_> =
+        found["matches"].as_array().unwrap().iter().map(|m| (m["start"]["off"].as_u64().unwrap(), m["end"]["off"].as_u64().unwrap())).collect();
+    assert_eq!(ranges, [(5, 11), (12, 20), (21, 29)]);
+    for options in [json!({"special": false}), json!({"wildcards": true}), json!({"regex": true})] {
+        let mut exact = super::s();
+        run(&mut exact, "document.setText", json!({"text": original}));
+        let mut params = options;
+        params["text"] = json!("Tau'va");
+        assert_eq!(run(&mut exact, "edit.find", params)["count"], 1);
+    }
+    for query in ["Tau^u39va", "Tau^u8217va", "Tau’va", "Tau‘va", "Tauʼva"] {
+        assert_eq!(run(&mut s, "edit.find", json!({"text": query}))["count"], 1, "{query}");
+    }
+    assert_eq!(run(&mut s, "edit.replaceAll", json!({"text": "Tau'va", "with": "<^&>"}))["replaced"], 3);
+    assert_eq!(text(&s), "🙂 <Tau'va> <Tau’va> <Tau‘va> Tauʼva");
+    run(&mut s, "edit.undo", json!({}));
+    assert_eq!(text(&s), original);
+}

crates/engine/src/tests/fonts.rs

@@ -469,3 +469,43 @@ fn an_east_asian_font_is_shown_as_the_face_its_text_is_drawn_in() {
     assert_eq!(m[0].shown_as, [shown], "{m:?}");
     assert!(crate::cmd::font::east_asian_shown_as(&doc, "JetBrains Mono", "JetBrains Mono").is_none(), "installed");
 }
+
+/// A semicolon-separated LibreOffice font fallback list is not reported as missing when any
+/// candidate is installed, and substitutes by candidate family when none is.
+#[test]
+fn libreoffice_font_fallback_lists_are_not_reported_as_missing_when_installed() {
+    let db = wordcraft_fonts::FontDb::global();
+    let doc = doc_of(vec![para_in("Text", "Liberation Sans;Arial")]);
+    if db.has_family("Liberation Sans") || db.has_family("Arial") {
+        let m = missing_in(doc.clone());
+        assert!(m.is_empty(), "expected not to be reported as missing: {m:?}");
+    }
+    let bundled_fallback = doc_of(vec![para_in("Text", "NonExistentSans123;Source Sans 3")]);
+    assert!(missing_in(bundled_fallback).is_empty(), "bundled fallback candidate is installed");
+
+    // Roundtrip through DOCX export and import:
+    let docx_bytes = crate::io::save_bytes("out.docx", &doc).unwrap();
+    let loaded = wordcraft_docx::read(&docx_bytes).unwrap();
+    assert_eq!(loaded.body[0].as_para().unwrap().runs[0].props.font.as_deref(), Some("Liberation Sans;Arial"));
+    if db.has_family("Liberation Sans") || db.has_family("Arial") {
+        let m = missing_in(loaded);
+        assert!(m.is_empty(), "expected not to report installed fallback font as missing: {m:?}");
+    }
+
+    // When none of the fonts is installed, missing_fonts flags it with an appropriate substitute:
+    if !db.has_family("NonExistentSans123") && !db.has_family("Times New Roman") {
+        let m = missing_in(doc_of(vec![para_in("Text", "NonExistentSans123;Times New Roman")]));
+        assert_eq!(m.len(), 1);
+        assert_eq!(m[0].font, "NonExistentSans123;Times New Roman");
+        let drawn = m[0].shown_as.first().expect("drawn face");
+        assert!(drawn.contains("Serif") || drawn == "Times" || drawn == "Tinos", "drawn: {drawn}");
+    }
+    let m = missing_in(doc_of(vec![para_in("Text", "NonExistentSans123;NonExistentSerif123")]));
+    assert_eq!(m.len(), 1);
+    assert_eq!(m[0].font, "NonExistentSans123;NonExistentSerif123");
+    let drawn = m[0].shown_as.first().expect("drawn face");
+    assert!(
+        drawn.contains("Serif") || drawn == "Times" || drawn == "Times New Roman" || drawn == "Tinos" || drawn == "Caladea" || drawn == "Georgia",
+        "drawn: {drawn}"
+    );
+}

crates/engine/src/tests/typing.rs

@@ -9,6 +9,19 @@ use wordcraft_doc::{Pos, StoryRef};
 
 use crate::Session;
 
+/// Leaf #152: deleting the empty heading above body text must not promote that text to a heading.
+/// Deleting the empty heading mark keeps the next paragraph as body text.
+#[test]
+fn deleting_a_preceding_empty_heading_keeps_the_body_paragraph_style() {
+    let mut s = Session::new(wordcraft_doc::Document::from_text("\nBody paragraph."));
+    run(&mut s, "para.style", json!({"style": "Heading1"}));
+    run(&mut s, "text.delete", json!({}));
+    let p = s.doc.para_at(&Pos::body(0, 0)).unwrap();
+    assert_eq!(p.text, "Body paragraph.");
+    assert_eq!(p.props.style, None);
+    assert_eq!(wordcraft_layout::outline::levels(&s.doc), [wordcraft_layout::outline::BODY]);
+}
+
 /// Create an empty session for command-driven typing fixtures.
 fn s() -> Session {
     Session::new(wordcraft_doc::Document::new())
@@ -1539,6 +1552,110 @@ fn typing_at_either_end_of_a_link_is_not_linked() {
     assert_eq!(links(&s), [("sixte".into(), false), (String::new(), false)]);
 }
 
+#[test]
+fn a_link_on_a_selection_shows_the_text_to_display() {
+    // Ctrl+K on selected text with a new Text to display: the selection is retyped with it and
+    // linked, keeping its formatting; typing after it stays outside the link (as in Word).
+    let mut s = typed("go see kiln now");
+    run(&mut s, "select.text", json!({"text": "kiln"}));
+    run(&mut s, "format.bold", json!({}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "Oven"}));
+    typ(&mut s, "!");
+    assert_eq!(texts(&s), ["go see oven! now"]);
+    assert_eq!(links(&s), [("Oven".into(), false)]);
+    let p = s.doc.body.iter().find_map(|b| b.as_para()).unwrap();
+    assert!(p.run_ranges().any(|(r, c)| &p.text[r] == "Oven" && c.bold == Some(true)), "the new text keeps the selection's formatting");
+    // The same text, or text that can't stand in one line, leaves the selection as it is.
+    run(&mut s, "select.text", json!({"text": "see"}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "see"}));
+    run(&mut s, "select.text", json!({"text": "now"}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "two\nlines"}));
+    assert_eq!(texts(&s), ["go see oven! now"]);
+}
+
+#[test]
+fn a_new_selection_drops_the_formatting_left_after_a_link() {
+    // A link leaves plain formatting for what is typed after it. A selection made next doesn't
+    // take it: retyping that selection keeps its own formatting.
+    let bold_run = |s: &Session, text: &str| {
+        let p = s.doc.body.iter().find_map(|b| b.as_para()).unwrap();
+        p.run_ranges().any(|(r, c)| &p.text[r] == text && c.bold == Some(true))
+    };
+    let mut s = typed("kiln");
+    run(&mut s, "select.text", json!({"text": "kiln"}));
+    run(&mut s, "format.bold", json!({}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "Oven"}));
+    run(&mut s, "select.all", json!({}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "Stove"}));
+    assert!(bold_run(&s, "Stove"), "select.all");
+    run(&mut s, "select.word", json!({}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "Range"}));
+    assert!(bold_run(&s, "Range"), "select.word");
+    let anchor = serde_json::to_value(s.doc.start_of(s.sel.focus.story)).unwrap();
+    let focus = serde_json::to_value(s.sel.focus.clone()).unwrap();
+    run(&mut s, "select.range", json!({"anchor": anchor, "focus": focus}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "Hob"}));
+    assert!(bold_run(&s, "Hob"), "select.range");
+}
+
+#[test]
+fn a_link_over_two_paragraphs_keeps_their_text() {
+    // Word shows <<Selection in Document>> for such a selection: the text to display is ignored.
+    let mut s = typed("one\ntwo");
+    run(&mut s, "select.all", json!({}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "both"}));
+    assert_eq!(texts(&s), ["one", "two"]);
+    assert_eq!(links(&s), [("One".into(), false), ("two".into(), false)]);
+}
+
+#[test]
+fn a_retyped_link_text_is_tracked() {
+    let mut s = typed("go see kiln now");
+    run(&mut s, "review.trackChanges", json!({"value": true}));
+    run(&mut s, "select.text", json!({"text": "kiln"}));
+    run(&mut s, "insert.link", json!({"url": "https://example.org/", "text": "Oven"}));
+    let p = s.doc.body.iter().find_map(|b| b.as_para()).unwrap();
+    assert!(p.run_ranges().any(|(r, c)| &p.text[r] == "kiln" && c.del.is_some()), "the old text stays as a deletion");
+    assert!(p.run_ranges().any(|(r, c)| &p.text[r] == "Oven" && c.ins.is_some() && c.link.is_some()));
+}
+
+#[test]
+fn a_link_keeps_a_selection_it_cannot_retype() {
+    use wordcraft_doc::props::CharProps;
+    use wordcraft_doc::{InlineObject, Path};
+    let url = "https://example.org/";
+    // A second retyped link keeps its own selection's formatting: the first one's left no
+    // formatting pending for it.
+    let mut s = typed("go see kiln and vat now");
+    run(&mut s, "select.text", json!({"text": "kiln"}));
+    run(&mut s, "insert.link", json!({"url": url, "text": "Oven"}));
+    run(&mut s, "select.text", json!({"text": "vat"}));
+    run(&mut s, "format.bold", json!({}));
+    run(&mut s, "insert.link", json!({"url": url, "text": "Tub"}));
+    let p = s.doc.body.iter().find_map(|b| b.as_para()).unwrap();
+    assert!(p.run_ranges().any(|(r, c)| &p.text[r] == "Tub" && c.bold == Some(true) && c.link.is_some()), "{:?}", p.text);
+    // A selection holding a manual line break keeps both lines.
+    let mut s = typed("one");
+    run(&mut s, "text.lineBreak", json!({}));
+    typ(&mut s, "two");
+    run(&mut s, "select.all", json!({}));
+    run(&mut s, "insert.link", json!({"url": url, "text": "both"}));
+    assert_eq!(s.doc.plain_text(StoryRef::Body), "one\ntwo");
+    // A selection in a field's result is only linked; the field's next update owns that text.
+    let mut doc = wordcraft_doc::Document::from_text("See ");
+    let c = CharProps::default();
+    let p = doc.para_mut(StoryRef::Body, &Path::top(0)).unwrap();
+    let start = p.len();
+    p.insert_object(start, InlineObject::FieldStart { instr: "ADDIN ZOTERO_ITEM CSL_CITATION {}".into(), locked: false, dirty: false }, &c).unwrap();
+    let n = p.insert_text(p.len(), "(Smith 2020)", &c).unwrap();
+    p.insert_object(start + 3 + n, InlineObject::FieldEnd, &c).unwrap();
+    let mut s = Session::new(doc);
+    run(&mut s, "select.text", json!({"text": "Smith"}));
+    run(&mut s, "insert.link", json!({"url": url, "text": "Jones"}));
+    assert_eq!(s.doc.plain_text(StoryRef::Body), "See (Smith 2020)");
+    assert_eq!(s.doc.field_ranges(StoryRef::Body).len(), 1);
+}
+
 #[test]
 fn typing_then_selecting_then_typing_are_two_undo_steps() {
     // #419: a selection that moves the insertion point ends the typing group, like a caret move.

crates/engine/src/view.rs

@@ -161,6 +161,17 @@ fn default_read_zoom() -> f32 {
     1.0
 }
 
+impl ViewState {
+    /// Take the app-wide settings a view keeps a copy of from the saved preferences: Read Mode's
+    /// page colour and Check spelling as you type (#5740). A view made active goes through this
+    /// (a document switched to, opened or made, a window's view put back), so a change made while
+    /// another view was active holds in this one too.
+    pub(crate) fn follow_prefs(&mut self, prefs: &crate::prefs::Prefs) {
+        self.read_page_color = prefs.read_page_color;
+        self.proofing = prefs.check_spelling;
+    }
+}
+
 impl Default for ViewState {
     fn default() -> Self {
         ViewState {

crates/engine/tests/empty_paragraph_merge.rs

@@ -0,0 +1,80 @@
+use serde_json::{Value, json};
+use wordcraft_doc::props::CharProps;
+use wordcraft_doc::{Document, Path, Pos, StoryRef};
+use wordcraft_engine::{Selection, Session};
+use wordcraft_layout::outline;
+
+/// Dispatch an engine command and report its ID if it fails.
+fn run(s: &mut Session, id: &str, params: Value) {
+    s.run(id, &params).unwrap_or_else(|e| panic!("{id}: {e}"));
+}
+
+/// Create an empty heading, mixed-run body text and a genuine TOC heading.
+fn fixture() -> Session {
+    let mut s = Session::new(Document::from_text("\nBody paragraph.\nReal heading\n"));
+    run(&mut s, "para.style", json!({"style": "Heading1"}));
+    s.doc.para_mut(StoryRef::Body, &Path::top(0)).unwrap().mark =
+        CharProps { font: Some("Source Sans 3".into()), size: Some(40.0), bold: Some(true), ..Default::default() };
+    let p = s.doc.para_mut(StoryRef::Body, &Path::top(1)).unwrap();
+    p.mark = CharProps { size: Some(11.0), bold: Some(false), ..Default::default() };
+    p.format(5, 14, &|c| c.italic = Some(true)).unwrap();
+    s.doc.para_mut(StoryRef::Body, &Path::top(2)).unwrap().props.style = Some("Heading2".into());
+    s
+}
+
+/// Keyboard and selection deletes preserve body formatting through Undo and Redo.
+#[test]
+fn empty_heading_delete_backspace_and_selection_preserve_body_format_and_history() {
+    for mode in ["delete", "backspace", "selection", "reverse selection"] {
+        for pending in [false, true] {
+            let mut s = fixture();
+            let expected = s.doc.para_at(&Pos::body(1, 0)).unwrap().clone();
+            s.sel = match mode {
+                "backspace" => Selection::caret(Pos::body(1, 0)),
+                "selection" => Selection { anchor: Pos::body(0, 0), focus: Pos::body(1, 0) },
+                "reverse selection" => Selection { anchor: Pos::body(1, 0), focus: Pos::body(0, 0) },
+                _ => Selection::caret(Pos::body(0, 0)),
+            };
+            if pending {
+                s.pending = Some(CharProps { size: Some(48.0), bold: Some(true), ..Default::default() });
+            }
+            let before = serde_json::to_value(&s.doc.body).unwrap();
+            run(&mut s, if mode == "backspace" { "text.backspace" } else { "text.delete" }, json!({}));
+            let p = s.doc.para_at(&Pos::body(0, 0)).unwrap();
+            assert_eq!(p.text, expected.text, "{mode}, pending={pending}");
+            assert_eq!(p.props, expected.props);
+            assert_eq!(p.mark, expected.mark);
+            assert_eq!(serde_json::to_value(&p.runs).unwrap(), serde_json::to_value(&expected.runs).unwrap());
+            let actual = s.doc.styles.resolve_char(p.props.style.as_deref(), p.props_of_char(0));
+            let old = s.doc.styles.resolve_char(expected.props.style.as_deref(), expected.props_of_char(0));
+            assert_eq!((actual.font, actual.size, actual.bold, actual.italic), (old.font, old.size, old.bold, old.italic));
+            assert!(s.pending.is_none());
+            assert_eq!(outline::levels(&s.doc)[..2], [outline::BODY, 2]);
+            let after = serde_json::to_value(&s.doc.body).unwrap();
+            run(&mut s, "edit.undo", json!({}));
+            assert_eq!(serde_json::to_value(&s.doc.body).unwrap(), before);
+            run(&mut s, "edit.redo", json!({}));
+            assert_eq!(serde_json::to_value(&s.doc.body).unwrap(), after);
+        }
+    }
+}
+
+/// Saving to DOCX and generating a TOC must keep the surviving body text out of the outline.
+#[test]
+fn deleted_empty_heading_does_not_promote_body_text_into_toc_or_docx() {
+    let mut s = fixture();
+    run(&mut s, "text.delete", json!({}));
+    assert_eq!(outline::levels(&s.doc)[..2], [outline::BODY, 2]);
+    let bytes = wordcraft_docx::write(&s.doc).unwrap();
+    let mut s = Session::new(wordcraft_docx::read(&bytes).unwrap());
+    assert_eq!(outline::levels(&s.doc)[..2], [outline::BODY, 2]);
+    let p = s.doc.para_at(&Pos::body(0, 0)).unwrap();
+    assert_eq!(p.props.style.as_deref(), None);
+    assert_eq!(p.mark.size, Some(11.0));
+    assert_eq!(p.props_of_char(6).italic, Some(true));
+    s.sel = Selection::caret(Pos::body(2, 0));
+    run(&mut s, "references.toc", json!({"levels": 3}));
+    let text = s.doc.plain_text(StoryRef::Body);
+    assert_eq!(text.matches("Body paragraph.").count(), 1, "body text is not a TOC entry: {text}");
+    assert_eq!(text.matches("Real heading").count(), 2, "the genuine heading stays in the TOC: {text}");
+}

crates/fonts/src/lib.rs

@@ -21,7 +21,7 @@ pub use fontdb::{
 pub use harfrust::Feature;
 use harfrust::{Direction, ShapeOptions, ShapePlan, ShapePlanKey, Shaper, Tag, UnicodeBuffer};
 pub use kurbo::BezPath;
-pub use script::{fallback_rank, is_emoji, needs_layout_tables, script_fallbacks};
+pub use script::{fallback_rank, is_emoji, joins_cursively, needs_layout_tables, script_fallbacks};
 use skrifa::MetadataProvider;
 use skrifa::instance::Size;
 pub use ui_fallback::{
@@ -185,13 +185,34 @@ fn direction_runs(text: &str) -> Vec<(std::ops::Range<usize>, bool)> {
 /// order: right-to-left runs are shaped right to left, then their clusters put back in text
 /// order (each cluster's glyphs keep the shaper's order); line layout reorders them visually.
 pub fn shape(face: &FontFace, text: &str, features: &[Feature], map: impl Fn(char) -> char) -> Vec<ShapedGlyph> {
+    shape_in(face, text, features, map, ShapeContext::default())
+}
+
+/// The text shown on either side of what is being shaped. Only a letter's form depends on it: the
+/// letters of an Arabic word join across a change of formatting, so each piece has to know what
+/// it continues (#4093). Nothing outside the shaped text is drawn or measured.
+#[derive(Clone, Copy, Debug, Default)]
+pub struct ShapeContext<'a> {
+    pub before: &'a str,
+    pub after: &'a str,
+}
+
+impl<'a> ShapeContext<'a> {
+    /// This context for the part `r` of `text`: what's beyond `text` where `r` reaches its edge.
+    fn at_edges_of(self, text: &str, r: &std::ops::Range<usize>) -> ShapeContext<'a> {
+        ShapeContext { before: if r.start == 0 { self.before } else { "" }, after: if r.end == text.len() { self.after } else { "" } }
+    }
+}
+
+/// [`shape`] of a piece of longer text (see [`ShapeContext`]).
+pub fn shape_in(face: &FontFace, text: &str, features: &[Feature], map: impl Fn(char) -> char, around: ShapeContext<'_>) -> Vec<ShapedGlyph> {
     if !text.chars().any(is_rtl) {
-        return shape_dir(face, text, features, &map, false);
+        return shape_dir(face, text, features, &map, false, around);
     }
     let mut out = Vec::with_capacity(text.len());
     for (r, rtl) in direction_runs(text) {
         let Some(sub) = text.get(r.clone()) else { continue };
-        let mut g = shape_run(face, sub, features, &map, rtl);
+        let mut g = shape_run_in(face, sub, features, &map, rtl, around.at_edges_of(text, &r));
         for x in &mut g {
             x.cluster += r.start;
         }
@@ -205,7 +226,19 @@ pub fn shape(face: &FontFace, text: &str, features: &[Feature], map: impl Fn(cha
 /// mark positioning. Glyphs come back in logical order like [`shape`]: clusters in text order,
 /// each cluster's glyphs in the shaper's (visual) order.
 pub fn shape_run(face: &FontFace, text: &str, features: &[Feature], map: impl Fn(char) -> char, rtl: bool) -> Vec<ShapedGlyph> {
-    let g = shape_dir(face, text, features, &map, rtl);
+    shape_run_in(face, text, features, map, rtl, ShapeContext::default())
+}
+
+/// [`shape_run`] of a piece of longer text (see [`ShapeContext`]).
+pub fn shape_run_in(
+    face: &FontFace,
+    text: &str,
+    features: &[Feature],
+    map: impl Fn(char) -> char,
+    rtl: bool,
+    around: ShapeContext<'_>,
+) -> Vec<ShapedGlyph> {
+    let g = shape_dir(face, text, features, &map, rtl, around);
     if !rtl {
         return g;
     }
@@ -221,7 +254,14 @@ pub fn shape_run(face: &FontFace, text: &str, features: &[Feature], map: impl Fn
     groups.into_iter().flatten().collect()
 }
 
-fn shape_dir(face: &FontFace, text: &str, features: &[Feature], map: &impl Fn(char) -> char, rtl: bool) -> Vec<ShapedGlyph> {
+fn shape_dir(
+    face: &FontFace,
+    text: &str,
+    features: &[Feature],
+    map: &impl Fn(char) -> char,
+    rtl: bool,
+    around: ShapeContext<'_>,
+) -> Vec<ShapedGlyph> {
     let mut out = Vec::with_capacity(text.len());
     let shaped = face.hb().map(|hb| {
         let shaper = face.shaper.shaper(&hb).instance(face.instance.as_ref()).build();
@@ -234,6 +274,10 @@ fn shape_dir(face: &FontFace, text: &str, features: &[Feature], map: &impl Fn(ch
             for (i, c) in sub.char_indices() {
                 buf.add(map(c), (r.start + i) as u32);
             }
+            // After the text: adding to a buffer clears what follows it.
+            let around = around.at_edges_of(text, &r);
+            buf.set_pre_context(around.before);
+            buf.set_post_context(around.after);
             buf.guess_segment_properties();
             buf.set_direction(if rtl { Direction::RightToLeft } else { Direction::LeftToRight });
             let plan = shape_plan(face, &shaper, &buf, features);
@@ -386,7 +430,10 @@ mod tests {
             gb.glyph_infos().iter().zip(gb.glyph_positions()).map(|(i, p)| (i.glyph_id, i.cluster, p.x_advance, p.x_offset, p.y_offset)).collect()
         };
         let kept = |face: &FontFace, text: &str, features: &[Feature], rtl: bool| -> Vec<(u32, u32, i32, i32, i32)> {
-            shape_dir(face, text, features, &|c| c, rtl).iter().map(|g| (g.gid, g.cluster as u32, g.x_advance, g.x_offset, g.y_offset)).collect()
+            shape_dir(face, text, features, &|c| c, rtl, ShapeContext::default())
+                .iter()
+                .map(|g| (g.gid, g.cluster as u32, g.x_advance, g.x_offset, g.y_offset))
+                .collect()
         };
         let db = FontDb::global();
         let faces = [db.face(DEFAULT_FAMILY, "Regular"), db.face(DEFAULT_FAMILY, "Italic"), db.face("Source Sans 3", "Regular")];
@@ -421,7 +468,8 @@ mod tests {
         for n in 0..3 * MAX_SHAPE_PLANS as u32 {
             let features = [feature_value(b"kern", n % 2), feature_value(b"liga", n)];
             let rtl = n % 3 == 0;
-            let kept: Vec<(u32, i32)> = shape_dir(&face, "office", &features, &|c| c, rtl).iter().map(|g| (g.gid, g.x_advance)).collect();
+            let kept: Vec<(u32, i32)> =
+                shape_dir(&face, "office", &features, &|c| c, rtl, ShapeContext::default()).iter().map(|g| (g.gid, g.x_advance)).collect();
             assert_eq!(kept, fresh(&features, rtl), "{n}");
             assert!(kept_shape_plans() <= MAX_SHAPE_PLANS, "{}", kept_shape_plans());
         }

crates/fonts/src/script.rs

@@ -18,6 +18,28 @@ fn real_script(c: char) -> Option<Script> {
     }
 }
 
+/// Whether `c` belongs to a script whose letters join their neighbours (Arabic, Syriac, Mongolian…),
+/// so a letter's form depends on the text on either side of it.
+pub fn joins_cursively(c: char) -> bool {
+    c >= '\u{0600}'
+        && matches!(
+            c.script(),
+            Script::Adlam
+                | Script::Arabic
+                | Script::Chorasmian
+                | Script::Hanifi_Rohingya
+                | Script::Mandaic
+                | Script::Manichaean
+                | Script::Mongolian
+                | Script::Nko
+                | Script::Old_Uyghur
+                | Script::Phags_Pa
+                | Script::Psalter_Pahlavi
+                | Script::Sogdian
+                | Script::Syriac
+        )
+}
+
 /// Maximal single-script byte ranges of `text`. Characters of no particular script join the run
 /// before them (or, at the start, the first run), so the ranges cover `text` exactly.
 pub(crate) fn script_runs(text: &str) -> Vec<std::ops::Range<usize>> {

crates/fonts/src/word.rs

@@ -32,12 +32,27 @@ fn substitutes(family: &str) -> &'static [&'static str] {
 /// as Liberation Serif for Times New Roman, or the same design under another name), rather than
 /// the generic lists any unknown font gets?
 pub fn has_named_substitutes(family: &str) -> bool {
+    if family.contains(';') {
+        return list_class(family).named;
+    }
     substitutes_by_name(charset_base_family(family)).1
 }
 
+const SANS: &[&str] = &["Carlito", "Helvetica Neue", "Arial", "Liberation Sans", "Noto Sans", "DejaVu Sans", "Source Sans 3"];
+
 /// [`substitutes`], and whether they were chosen for this family by name.
 fn substitutes_by_name(family: &str) -> (&'static [&'static str], bool) {
-    const SANS: &[&str] = &["Carlito", "Helvetica Neue", "Arial", "Liberation Sans", "Noto Sans", "DejaVu Sans", "Source Sans 3"];
+    if family.contains(';') {
+        for cand in family.split(';').map(str::trim).filter(|s| !s.is_empty()) {
+            let (subs, named) = substitutes_by_name(cand);
+            if named || is_thai_family(cand) || subs != SANS {
+                return (subs, named);
+            }
+        }
+        if let Some(first) = family.split(';').map(str::trim).find(|s| !s.is_empty()) {
+            return substitutes_by_name(first);
+        }
+    }
     const SERIF: &[&str] = &["Caladea", "Times New Roman", "Liberation Serif", "Georgia", "Noto Serif", "DejaVu Serif", "Source Serif 4"];
     const MONO: &[&str] = &["Courier New", "Liberation Mono", "Menlo", "Consolas", "DejaVu Sans Mono", "JetBrains Mono"];
     const ARIAL: &[&str] = &["Liberation Sans", "Helvetica", "Helvetica Neue", "Arimo", "Source Sans 3"];
@@ -157,11 +172,11 @@ fn substitutes_by_name(family: &str) -> (&'static [&'static str], bool) {
 /// itself: a Symbol bullet makes its line taller than the text alone. The values are those Word's
 /// PDFs give the fonts (`/Ascent`, `/Descent`).
 pub fn symbol_font_metrics(family: &str) -> Option<(f64, f64)> {
-    match family.trim().to_ascii_lowercase().as_str() {
+    family.split(';').map(str::trim).filter(|f| !f.is_empty()).find_map(|f| match f.to_ascii_lowercase().as_str() {
         "symbol" | "symbol (pcl6)" => Some((1.005, 0.216)),
         "wingdings" => Some((0.899, 0.205)),
         _ => None,
-    }
+    })
 }
 
 /// The (ascent, descent) in em Word gives a line with a run in a symbol font, when that isn't the
@@ -224,7 +239,10 @@ fn metric_twins(family: &str) -> impl Iterator<Item = &'static str> + '_ {
 /// wraps and paginates as in `family`.
 pub fn metric_compatible(family: &str, substitute: &str) -> bool {
     let s = substitute.trim();
-    metric_twins(charset_base_family(family.trim())).any(|t| t.eq_ignore_ascii_case(s))
+    family.split(';').map(str::trim).filter(|f| !f.is_empty()).any(|f| {
+        let base = charset_base_family(f);
+        base.eq_ignore_ascii_case(s) || metric_twins(base).any(|t| t.eq_ignore_ascii_case(s))
+    })
 }
 
 /// Naskh-style Persian/Arabic document fonts (the "B …" family names come from the common
@@ -356,10 +374,63 @@ fn is_thai_serif(f: &str) -> bool {
 /// Is `family` a Thai document font ([`is_thai`])? A missing one is drawn in a Thai face for the
 /// whole run, never in a Latin substitute (#1549).
 pub fn is_thai_family(family: &str) -> bool {
-    let f = family.to_ascii_lowercase();
+    if family.contains(';') {
+        return list_class(family).thai;
+    }
+    let f = family.trim().to_ascii_lowercase();
     is_thai(&f) || is_thai_serif(&f)
 }
 
+/// What the name-based predicates answer about one semicolon-separated fallback list. Shaping asks
+/// them for every piece of every run set in the list, so the answer is worked out once per list
+/// rather than once per piece.
+#[derive(Clone, Copy, Default)]
+struct ListClass {
+    /// [`is_thai_family`].
+    thai: bool,
+    /// [`has_named_substitutes`].
+    named: bool,
+}
+
+/// [`ListClass`] of the semicolon-separated `family`, memoised by its text.
+fn list_class(family: &str) -> ListClass {
+    let key = family_list_key(family);
+    let generation = FontDb::global().generation();
+    {
+        let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+        if c.generation != generation {
+            *c = FallbackListMemo { generation, ..Default::default() };
+        } else if let Some((cached, class)) = c.classes.get(&key)
+            && cached == family
+        {
+            return *class;
+        }
+    }
+    let class = ListClass { thai: classify_thai(family), named: classify_named(family) };
+    let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+    if c.generation == generation {
+        if c.classes.len() >= FAMILY_CACHE_MAX {
+            c.classes.clear();
+        }
+        c.classes.insert(key, (family.to_string(), class));
+    }
+    class
+}
+
+/// Whether any family of the list is a Thai document font: what [`is_thai_family`] answers.
+fn classify_thai(family: &str) -> bool {
+    family.split(';').map(str::trim).filter(|f| !f.is_empty()).any(|f| {
+        let f = f.to_ascii_lowercase();
+        is_thai(&f) || is_thai_serif(&f)
+    })
+}
+
+/// Whether any family of the list has substitutes of its own: what [`has_named_substitutes`]
+/// answers.
+fn classify_named(family: &str) -> bool {
+    family.split(';').map(str::trim).filter(|f| !f.is_empty()).any(|f| substitutes_by_name(charset_base_family(f)).1)
+}
+
 /// Khmer document fonts: the Khmer OS family, Khmer Mool1 and the like, Windows' and the Mac's
 /// own and Noto's. Matched by name only.
 fn is_khmer(f: &str) -> bool {
@@ -416,8 +487,168 @@ fn resolve_cache() -> &'static Mutex<Memo> {
 }
 
 const FAMILY_CACHE_MAX: usize = 4096;
-/// Family names longer than this (bytes) are resolved but never remembered.
+/// Family names longer than this (bytes) are resolved but never remembered as string keys.
 const FAMILY_KEY_MAX: usize = 256;
+/// Maximum number of direct `FontDb::has_family` catalog scans per uncached fallback list
+/// for candidate names not present in [`FontDb::families`].
+const MAX_DIRECT_CANDIDATE_PROBES: usize = 32;
+
+/// Memoised fallback-list resolutions and installed-family lookups for the current [`FontDb::generation`].
+#[derive(Default)]
+struct FallbackListMemo {
+    generation: u64,
+    families: std::collections::HashSet<String>,
+    probed: HashMap<String, bool>,
+    lists: HashMap<(u64, usize), (String, (String, bool))>,
+    long_resolved: HashMap<(u64, usize, bool, bool), (String, Resolved)>,
+    classes: HashMap<(u64, usize), (String, ListClass)>,
+}
+
+/// Shared cache for fallback-list resolutions and candidate availability checks.
+fn fallback_list_cache() -> &'static Mutex<FallbackListMemo> {
+    static CACHE: OnceLock<Mutex<FallbackListMemo>> = OnceLock::new();
+    CACHE.get_or_init(|| Mutex::new(FallbackListMemo::default()))
+}
+
+/// Hash and byte length of `family`, used as a compact cache key for fallback lists.
+fn family_list_key(family: &str) -> (u64, usize) {
+    use std::hash::{Hash, Hasher};
+    let mut h = std::collections::hash_map::DefaultHasher::new();
+    family.hash(&mut h);
+    (h.finish(), family.len())
+}
+
+/// Whether `cand` is an installed family in `db`, using the memoised family set and bounded
+/// direct catalog probes so long fallback lists do not rescan the font catalog for every candidate.
+fn candidate_has_family(db: &FontDb, cand: &str, direct_probes: &mut usize) -> bool {
+    let cand = cand.trim();
+    if cand.is_empty() || cand.len() > FAMILY_KEY_MAX {
+        return false;
+    }
+    let generation = db.generation();
+    let lower = cand.to_ascii_lowercase();
+    {
+        let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+        if c.generation != generation {
+            *c = FallbackListMemo { generation, ..Default::default() };
+        }
+        if c.families.is_empty() {
+            c.families = db.families().into_iter().map(|f| f.to_ascii_lowercase()).collect();
+        }
+        if c.families.contains(&lower) {
+            return true;
+        }
+        if let Some(&hit) = c.probed.get(&lower) {
+            return hit;
+        }
+    }
+    if *direct_probes == 0 {
+        return false;
+    }
+    *direct_probes -= 1;
+    let hit = db.has_family(cand);
+    let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+    if c.generation == generation {
+        if c.probed.len() >= FAMILY_CACHE_MAX {
+            c.probed.clear();
+        }
+        c.probed.insert(lower, hit);
+    }
+    hit
+}
+
+/// Record a resolved `(family, substituted)` pair for a semicolon-separated fallback list.
+fn remember_list_resolution(generation: u64, key: (u64, usize), family: &str, result: (String, bool)) -> (String, bool) {
+    let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+    if c.generation == generation {
+        if c.lists.len() >= FAMILY_CACHE_MAX {
+            c.lists.clear();
+        }
+        c.lists.insert(key, (family.to_string(), result.clone()));
+    }
+    result
+}
+
+/// Resolve a single family or semicolon-separated fallback list to `(family_name, substituted)`.
+fn resolve_family(db: &FontDb, family: &str) -> (String, bool) {
+    if !family.contains(';') {
+        let base = if family.trim().eq_ignore_ascii_case("Symbol (PCL6)") { "Symbol" } else { charset_base_family(family) };
+        return if db.has_family(family) {
+            (family.to_string(), false)
+        } else if base != family && db.has_family(base) {
+            (base.to_string(), false)
+        } else {
+            (substitute(db, base).unwrap_or(crate::FALLBACK_FAMILY).to_string(), true)
+        };
+    }
+
+    let generation = db.generation();
+    let list_key = family_list_key(family);
+    {
+        let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+        if c.generation != generation {
+            *c = FallbackListMemo { generation, ..Default::default() };
+        } else if let Some((cached_family, r)) = c.lists.get(&list_key)
+            && cached_family == family
+        {
+            return r.clone();
+        }
+    }
+
+    let candidates: Vec<&str> = family.split(';').map(str::trim).filter(|s| !s.is_empty() && s.len() <= FAMILY_KEY_MAX).collect();
+    if candidates.is_empty() {
+        return remember_list_resolution(generation, list_key, family, (crate::FALLBACK_FAMILY.to_string(), true));
+    }
+
+    let mut direct_probes = MAX_DIRECT_CANDIDATE_PROBES;
+
+    // 1. First installed family in the list:
+    for &cand in &candidates {
+        let base = if cand.eq_ignore_ascii_case("Symbol (PCL6)") { "Symbol" } else { charset_base_family(cand) };
+        if candidate_has_family(db, cand, &mut direct_probes) {
+            return remember_list_resolution(generation, list_key, family, (cand.to_string(), false));
+        }
+        if base != cand && candidate_has_family(db, base, &mut direct_probes) {
+            return remember_list_resolution(generation, list_key, family, (base.to_string(), false));
+        }
+    }
+
+    // 2. None installed: fall back to an installed substitute for an entry in the list rather than generic Carlito.
+    // Try Thai substitutes before metric twins of later candidates:
+    for &cand in &candidates {
+        let base = if cand.eq_ignore_ascii_case("Symbol (PCL6)") { "Symbol" } else { charset_base_family(cand) };
+        if is_thai_family(base)
+            && let Some(sub) = substitute(db, base)
+        {
+            return remember_list_resolution(generation, list_key, family, (sub.to_string(), true));
+        }
+        let mut twin_probes = MAX_DIRECT_CANDIDATE_PROBES;
+        if let Some(twin) = metric_twins(base).find(|t| candidate_has_family(db, t, &mut twin_probes)) {
+            return remember_list_resolution(generation, list_key, family, (twin.to_string(), true));
+        }
+    }
+
+    // Try candidates with named or category substitutes (Thai, Hebrew, Khmer, Persian, serif, mono):
+    for &cand in &candidates {
+        let base = if cand.eq_ignore_ascii_case("Symbol (PCL6)") { "Symbol" } else { charset_base_family(cand) };
+        let (subs, named) = substitutes_by_name(base);
+        if (named || is_thai_family(base) || subs != SANS)
+            && let Some(sub) = substitute(db, base)
+        {
+            return remember_list_resolution(generation, list_key, family, (sub.to_string(), true));
+        }
+    }
+
+    // Fall back to the first candidate's substitute:
+    if let Some(&first) = candidates.first() {
+        let base = if first.eq_ignore_ascii_case("Symbol (PCL6)") { "Symbol" } else { charset_base_family(first) };
+        if let Some(sub) = substitute(db, base) {
+            return remember_list_resolution(generation, list_key, family, (sub.to_string(), true));
+        }
+    }
+
+    remember_list_resolution(generation, list_key, family, (crate::FALLBACK_FAMILY.to_string(), true))
+}
 
 /// Resolve `family` with bold/italic to a face (memoised until user fonts are added or removed,
 /// [`FontDb::generation`]).
@@ -427,6 +658,10 @@ pub fn resolve(family: &str, bold: bool, italic: bool) -> Resolved {
     let generation = db.generation();
     let cacheable = family.len() <= FAMILY_KEY_MAX;
     let key = (if cacheable { family.to_string() } else { String::new() }, bold, italic);
+    let long_key = (!cacheable && family.contains(';')).then(|| {
+        let (h, len) = family_list_key(family);
+        (h, len, bold, italic)
+    });
     {
         let mut c = cache.lock().unwrap_or_else(std::sync::PoisonError::into_inner);
         if c.0 != generation {
@@ -436,24 +671,35 @@ pub fn resolve(family: &str, bold: bool, italic: bool) -> Resolved {
             return *r;
         }
     }
-    // Printer-driver Symbol aliases use the same encoding as Symbol.
-    let base = if family.trim().eq_ignore_ascii_case("Symbol (PCL6)") { "Symbol" } else { charset_base_family(family) };
-    let (fam, substituted) = if db.has_family(family) {
-        (family.to_string(), false)
-    } else if base != family && db.has_family(base) {
-        // Windows maps the charset names to the font itself: not a substitution.
-        (base.to_string(), false)
-    } else {
-        (substitute(db, base).unwrap_or(crate::FALLBACK_FAMILY).to_string(), true)
-    };
+    if let Some(lk) = long_key {
+        let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+        if c.generation != generation {
+            *c = FallbackListMemo { generation, ..Default::default() };
+        } else if let Some((cached_family, r)) = c.long_resolved.get(&lk)
+            && cached_family == family
+        {
+            return *r;
+        }
+    }
+    let (fam, substituted) = resolve_family(db, family);
     let face = db.face(&fam, style_name(bold, italic));
     let r = Resolved { synth_bold: bold && face.weight < 600.0, synth_italic: italic && !face.italic, substituted, face: FaceRef::of(&face) };
-    let mut c = cache.lock().unwrap_or_else(std::sync::PoisonError::into_inner);
-    if cacheable && c.0 == generation {
-        if c.1.len() >= FAMILY_CACHE_MAX {
-            c.1.clear();
+    if cacheable {
+        let mut c = cache.lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+        if c.0 == generation {
+            if c.1.len() >= FAMILY_CACHE_MAX {
+                c.1.clear();
+            }
+            c.1.insert(key, r);
+        }
+    } else if let Some(lk) = long_key {
+        let mut c = fallback_list_cache().lock().unwrap_or_else(std::sync::PoisonError::into_inner);
+        if c.generation == generation {
+            if c.long_resolved.len() >= FAMILY_CACHE_MAX {
+                c.long_resolved.clear();
+            }
+            c.long_resolved.insert(lk, (family.to_string(), r));
         }
-        c.1.insert(key, r);
     }
     r
 }
@@ -787,4 +1033,114 @@ mod tests {
             assert!(!has_named_substitutes(f), "{f}");
         }
     }
+
+    /// Semicolon-separated fallback lists resolve to the first installed family, or to a
+    /// candidate's substitute when no listed family is installed.
+    #[test]
+    fn libreoffice_fallback_list_resolves_first_installed_family() {
+        let db = FontDb::global();
+        if db.has_family("Liberation Sans") || db.has_family("Arial") {
+            let r = resolve("Liberation Sans;Arial", false, false);
+            assert!(!r.substituted, "should not be substituted when a candidate is installed: got {:?}", r.face.family);
+            if db.has_family("Liberation Sans") {
+                assert_eq!(r.face.family, "Liberation Sans");
+            } else {
+                assert_eq!(r.face.family, "Arial");
+            }
+        }
+        let bundled = resolve("NonExistentSansFace123;Source Serif 4", false, false);
+        assert!(!bundled.substituted);
+        assert_eq!(bundled.face.family, "Source Serif 4");
+
+        // When none of the families is installed, fall back to a substitute for an entry in the list rather than generic Carlito:
+        if !db.has_family("NonExistentSansFace123") && !db.has_family("Times New Roman") {
+            let r = resolve("NonExistentSansFace123;Times New Roman", false, false);
+            assert!(r.substituted);
+            assert_ne!(r.face.family, "Carlito", "should not fall back to generic Carlito when an entry is Times New Roman");
+            assert!(r.face.family.contains("Serif") || r.face.family == "Times" || r.face.family == "Tinos", "got {:?}", r.face.family);
+        }
+        let r = resolve("NonExistentSansFace123;NonExistentSerifFace123", false, false);
+        assert!(r.substituted);
+        assert_ne!(r.face.family, "Carlito", "should not fall back to generic Carlito when an entry is a serif family");
+        assert!(
+            r.face.family.contains("Serif")
+                || r.face.family == "Times"
+                || r.face.family == "Times New Roman"
+                || r.face.family == "Tinos"
+                || r.face.family == "Caladea"
+                || r.face.family == "Georgia",
+            "got {:?}",
+            r.face.family
+        );
+
+        // Metric compatibility of fallback list:
+        assert!(metric_compatible("Open Sans;Arial", "Liberation Sans"));
+        assert!(metric_compatible("Liberation Sans;Arial", "Helvetica"));
+        assert!(!metric_compatible("Open Sans;Arial", "Carlito"));
+    }
+
+    /// In a mixed Thai-and-Latin fallback list where neither listed family is installed, an
+    /// earlier Thai substitute takes precedence over a later candidate's metric twin.
+    #[test]
+    fn thai_fallback_list_prefers_thai_substitute_before_later_metric_twin() {
+        let db = FontDb::global();
+        let Some(thai_sub) = substitute(db, "NonExistentThaiUPC") else {
+            eprintln!("skipped: no Thai substitute installed");
+            return;
+        };
+        if !db.has_family("Cordia New") && !db.has_family("Arial") && metric_twins("Arial").any(|t| db.has_family(t)) {
+            let r = resolve("Cordia New;Arial", false, false);
+            assert!(r.substituted);
+            assert!(substitutes("Cordia New").contains(&r.face.family.as_str()), "got {:?}", r.face.family);
+        }
+        if let Some(latin) =
+            METRIC_TWINS.iter().flat_map(|g| g.iter().copied()).find(|&f| !db.has_family(f) && metric_twins(f).any(|t| db.has_family(t)))
+        {
+            let list = format!("NonExistentThaiUPC;{latin}");
+            let r = resolve(&list, false, false);
+            assert!(r.substituted);
+            assert_eq!(r.face.family, thai_sub);
+        }
+    }
+
+    /// Fallback lists longer than `FAMILY_KEY_MAX` still select a later installed family and
+    /// cache their resolution without growing `resolve_cache` string keys.
+    #[test]
+    fn long_fallback_list_selects_later_installed_family_and_caches() {
+        let mut candidates: Vec<String> = (0..120).map(|i| format!("MissingCandidateFont{i}")).collect();
+        candidates.push("Source Serif 4".to_string());
+        let list = candidates.join(";");
+        assert!(list.len() > FAMILY_KEY_MAX);
+
+        let r1 = resolve(&list, false, false);
+        let r2 = resolve(&list, false, false);
+        assert!(!r1.substituted);
+        assert_eq!(r1.face.family, "Source Serif 4");
+        assert_eq!(r1.face.id(), r2.face.id());
+        assert!(!resolve_cache().lock().unwrap_or_else(|e| e.into_inner()).1.keys().any(|k| k.0.len() > FAMILY_KEY_MAX));
+    }
+
+    /// The name-based predicates look at every candidate, and their answer is cached per list: a
+    /// later Thai or named fallback still counts, and a list is classified once rather than for
+    /// every piece of text shaped in it.
+    #[test]
+    fn list_classification_reads_every_candidate_and_is_cached() {
+        // The Thai family is last, past any shorter bound, and still classifies the list.
+        let mut candidates: Vec<String> = (0..200).map(|i| format!("MissingCandidateFont{i}")).collect();
+        candidates.push("TH Sarabun New".to_string());
+        let thai_list = candidates.join(";");
+        assert!(is_thai_family(&thai_list), "a later Thai family still counts");
+
+        // A named substitute at the end classifies too (Times New Roman has Liberation Serif).
+        candidates.pop();
+        candidates.push("Times New Roman".to_string());
+        let named_list = candidates.join(";");
+        assert!(has_named_substitutes(&named_list), "a later named fallback still counts");
+
+        // Both are memoised by list text: a second ask answers without rescanning, and a list
+        // whose text differs only after the key's hash is not mistaken for the cached one.
+        assert!(is_thai_family(&thai_list) && !is_thai_family(&named_list));
+        let cached = fallback_list_cache().lock().unwrap_or_else(|e| e.into_inner()).classes.len();
+        assert!(cached >= 2, "{cached} lists classified");
+    }
 }

crates/formats/src/html/entity.rs

@@ -3,6 +3,10 @@
 
 /// Decode an entity body (`amp`, `#38`, `#x26`) to its text.
 pub fn decode_entity(name: &str) -> Option<String> {
+    decode_char(name).map(String::from)
+}
+
+fn decode_char(name: &str) -> Option<char> {
     if let Some(num) = name.strip_prefix('#') {
         let (digits, radix) = match num.strip_prefix(['x', 'X']) {
             Some(h) => (h, 16),
@@ -17,7 +21,7 @@ pub fn decode_entity(name: &str) -> Option<String> {
             0x80..=0x9F => crate::txt::cp1252(v as u8),
             _ => char::from_u32(v).unwrap_or('\u{FFFD}'),
         };
-        return Some(c.to_string());
+        return Some(c);
     }
     let c = match name {
         "amp" | "AMP" => '&',
@@ -229,11 +233,54 @@ pub fn decode_entity(name: &str) -> Option<String> {
         "zwnj" => '\u{200C}',
         _ => return None,
     };
-    Some(c.to_string())
+    Some(c)
 }
 
-/// Replace entity references in `s`.
+/// Legacy names HTML5 also decodes without a `;` (only those [`decode_char`] knows), sorted by bytes.
+const BARE_NAMES: &[&str] = &[
+    "AElig", "AMP", "Aacute", "Acirc", "Agrave", "Aring", "Atilde", "Auml", "COPY", "Ccedil", "Eacute", "Ecirc", "Egrave", "Euml", "GT", "Iacute",
+    "Icirc", "Igrave", "Iuml", "LT", "Ntilde", "Oacute", "Ocirc", "Ograve", "Oslash", "Otilde", "Ouml", "QUOT", "REG", "Uacute", "Ucirc", "Ugrave",
+    "Uuml", "Yacute", "aacute", "acirc", "aelig", "agrave", "amp", "aring", "atilde", "auml", "ccedil", "cent", "copy", "curren", "deg", "divide",
+    "eacute", "ecirc", "egrave", "euml", "frac12", "frac14", "frac34", "gt", "iacute", "icirc", "iexcl", "igrave", "iquest", "iuml", "laquo", "lt",
+    "micro", "middot", "nbsp", "ntilde", "oacute", "ocirc", "ograve", "oslash", "otilde", "ouml", "para", "plusmn", "pound", "quot", "raquo", "reg",
+    "sect", "shy", "sup1", "sup2", "sup3", "szlig", "times", "uacute", "ucirc", "ugrave", "uuml", "yacute", "yen", "yuml",
+];
+
+/// A reference written without its `;` at the start of `after` (the text after `&`): a numeric one,
+/// or the longest legacy name. In attribute values a name followed by `=` or an alphanumeric stays text.
+fn bare_reference(after: &str, attr: bool) -> Option<(char, usize)> {
+    if let Some(num) = after.strip_prefix('#') {
+        let hex = num.starts_with(['x', 'X']);
+        let radix = if hex { 16 } else { 10 };
+        let skip = usize::from(hex);
+        let n = num.get(skip..)?.bytes().take_while(|b| char::from(*b).is_digit(radix)).count();
+        if n == 0 {
+            return None;
+        }
+        let len = 1 + skip + n;
+        return decode_char(after.get(..len)?).map(|c| (c, len));
+    }
+    // Legacy names are 2 to 6 ASCII alphanumerics: try the longest prefix first.
+    let run = after.bytes().take(6).take_while(u8::is_ascii_alphanumeric).count();
+    let len = (2..=run).rev().find(|&l| after.get(..l).is_some_and(|p| BARE_NAMES.binary_search(&p).is_ok()))?;
+    let next = after.as_bytes().get(len);
+    if attr && next.is_some_and(|b| b.is_ascii_alphanumeric() || *b == b'=') {
+        return None;
+    }
+    decode_char(after.get(..len)?).map(|c| (c, len))
+}
+
+/// Replace entity references in text `s`.
 pub fn unescape(s: &str) -> String {
+    unescape_in(s, false)
+}
+
+/// Replace entity references in an attribute value `s`.
+pub fn unescape_attr(s: &str) -> String {
+    unescape_in(s, true)
+}
+
+fn unescape_in(s: &str, attr: bool) -> String {
     if !s.contains('&') {
         return s.to_string();
     }
@@ -248,10 +295,10 @@ pub fn unescape(s: &str) -> String {
             cut -= 1;
         }
         let end = after.get(..cut).and_then(|w| w.find(';'));
-        match end.and_then(|e| decode_entity(&after[..e]).map(|t| (t, e))) {
-            Some((t, e)) => {
-                out.push_str(&t);
-                rest = &after[e + 1..];
+        match end.and_then(|e| decode_char(&after[..e]).map(|c| (c, e + 1))).or_else(|| bare_reference(after, attr)) {
+            Some((c, e)) => {
+                out.push(c);
+                rest = &after[e..];
             }
             None => {
                 out.push('&');
@@ -310,3 +357,14 @@ pub fn decode_as(b: &[u8], charset: Option<&str>) -> String {
         },
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn bare_names_are_sorted_and_known() {
+        assert!(BARE_NAMES.windows(2).all(|w| w[0] < w[1]));
+        assert!(BARE_NAMES.iter().all(|n| (2..=6).contains(&n.len()) && decode_char(n).is_some()));
+    }
+}

crates/formats/src/html/lexer.rs

@@ -1,4 +1,4 @@
-use super::entity::unescape;
+use super::entity::{unescape, unescape_attr};
 
 // ---------------------------------------------------------------------------------------------
 // Tokenizer
@@ -154,7 +154,7 @@ impl Lexer<'_> {
                     Some(&q) if q == b'"' || q == b'\'' => {
                         let v0 = i + 1;
                         let e = rest.get(v0..).and_then(|r| r.find(q as char)).map(|k| v0 + k).unwrap_or(rest.len());
-                        val = unescape(rest.get(v0..e).unwrap_or(""));
+                        val = unescape_attr(rest.get(v0..e).unwrap_or(""));
                         i = e + 1;
                     }
                     _ => {
@@ -162,7 +162,7 @@ impl Lexer<'_> {
                         while b.get(i).is_some_and(|c| !c.is_ascii_whitespace() && *c != b'>') {
                             i += 1;
                         }
-                        val = unescape(rest.get(v0..i).unwrap_or(""));
+                        val = unescape_attr(rest.get(v0..i).unwrap_or(""));
                     }
                 }
             }

crates/formats/src/html/tests.rs

@@ -232,6 +232,21 @@ fn entities() {
     assert_eq!(unescape("θ λ Δ ∇ ∑ √ ≠ ≤ ≥ ♥"), "θ λ Δ ∇ ∑ √ ≠ ≤ ≥ ♥");
 }
 
+/// Browsers decode numeric and legacy named references without their `;` in text,
+/// but keep `&name=` and `&nameX` intact inside attribute values (query strings).
+#[test]
+fn references_without_semicolon() {
+    assert_eq!(unescape("&copy 2024 Acme"), "© 2024 Acme");
+    assert_eq!(unescape("&#169 2024 Acme"), "© 2024 Acme");
+    assert_eq!(unescape("&#xA9 2024 &amp &lt"), "© 2024 & <");
+    assert_eq!(unescape("&bogus &# &#x &"), "&bogus &# &#x &");
+    assert_eq!(unescape("&nbsp;x"), "\u{00A0}x");
+    let f = parse("<p><a href=\"/q?a=1&copy=2&reg2=3&amp;b=&lt\">l</a> &copy 2024</p>");
+    let FBlock::Para(p) = &f.blocks[0] else { panic!() };
+    assert!(p.inlines.iter().any(|i| matches!(i, Inline::Text(t, f) if t == "l" && f.link.as_deref() == Some("/q?a=1&copy=2&reg2=3&b=<"))));
+    assert!(p.text().ends_with("© 2024"));
+}
+
 #[test]
 fn wbr_word_break_opportunity() {
     let f = parse("<p>Super<wbr>cali<wbr>fragilistic</p>");