mirror of
https://github.com/storytold/gridcraft.git
synced 2026-10-24 18:56:59 -06:00
Compare commits
1
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
 Dylan M. Taylor
|
a2f630f043 |
@@ -248,11 +248,10 @@ impl Discovery {
|
||||
/// (`~$…`) are skipped.
|
||||
fn collect_inputs(inputs: &[String]) -> Discovery {
|
||||
let mut out = Discovery::default();
|
||||
let mut seen = std::collections::HashSet::new();
|
||||
for i in inputs {
|
||||
let p = PathBuf::from(i);
|
||||
if p.is_dir() {
|
||||
if !walk(&p, 0, &mut out, &mut seen) {
|
||||
if !walk(&p, 0, &mut out) {
|
||||
break;
|
||||
}
|
||||
} else if !out.add_file(p) {
|
||||
@@ -269,20 +268,7 @@ fn collect_inputs(inputs: &[String]) -> Discovery {
|
||||
}
|
||||
|
||||
/// Returns false only when the global file budget stops traversal.
|
||||
fn walk(dir: &Path, depth: usize, out: &mut Discovery, seen: &mut std::collections::HashSet<u64>) -> bool {
|
||||
// A directory is visited once by identity (dev+inode on unix): several links to the same
|
||||
// directory are aliases, not exponential branches of the scan. Other platforms keep the
|
||||
// depth cap only.
|
||||
#[cfg(unix)]
|
||||
{
|
||||
use std::os::unix::fs::MetadataExt as _;
|
||||
if let Ok(m) = std::fs::metadata(dir)
|
||||
&& let Some(id) = m.dev().checked_mul(1 << 60).and_then(|d| d.checked_add(m.ino()))
|
||||
&& !seen.insert(id)
|
||||
{
|
||||
return true;
|
||||
}
|
||||
}
|
||||
fn walk(dir: &Path, depth: usize, out: &mut Discovery) -> bool {
|
||||
if depth > MAX_DEPTH {
|
||||
out.fail(dir, format!("directory depth limit of {MAX_DEPTH} exceeded; this subtree was not checked"));
|
||||
return true;
|
||||
@@ -315,7 +301,7 @@ fn walk(dir: &Path, depth: usize, out: &mut Discovery, seen: &mut std::collectio
|
||||
}
|
||||
};
|
||||
if metadata.is_dir() {
|
||||
if !walk(&p, depth + 1, out, seen) {
|
||||
if !walk(&p, depth + 1, out) {
|
||||
return false;
|
||||
}
|
||||
} else if p.extension().and_then(|e| e.to_str()).is_some_and(|e| e.eq_ignore_ascii_case("xlsx") || e.eq_ignore_ascii_case("xlsm"))
|
||||
@@ -995,12 +981,6 @@ fn relationship_kinds(bytes: &[u8]) -> Result<BTreeMap<String, String>, String>
|
||||
let mut zip = zip::ZipArchive::new(Cursor::new(bytes)).map_err(|e| format!("not a zip package: {e}"))?;
|
||||
let names = zip_names(&mut zip);
|
||||
let by_key: HashMap<String, &String> = names.iter().map(|n| (part_key(n), n)).collect();
|
||||
// Several relationships name the same part; each fresh check decompressed it again, so
|
||||
// repeated references to one highly compressible part multiplied the work. The verdict is
|
||||
// remembered per (part, named content) and the decompression the package can cause is
|
||||
// budgeted as a whole (finding: repeated decompression).
|
||||
let mut verdicts: HashMap<(String, &'static str), bool> = HashMap::new();
|
||||
let mut budget: u64 = 512 * 1024 * 1024;
|
||||
let mut kinds = BTreeMap::new();
|
||||
for n in names.iter().filter(|n| part_key(n).ends_with(".rels")) {
|
||||
let Some(src_dir) = rels_source_dir(&part_key(n)) else { continue };
|
||||
@@ -1014,7 +994,7 @@ fn relationship_kinds(bytes: &[u8]) -> Result<BTreeMap<String, String>, String>
|
||||
let external = attr(attrs, "TargetMode").is_some_and(|m| m.eq_ignore_ascii_case("External"));
|
||||
if !external
|
||||
&& let Some(part) = by_key.get(&resolve_target(&src_dir, target).to_ascii_lowercase())
|
||||
&& !holds_something(&mut zip, part, kind, &mut verdicts, &mut budget)
|
||||
&& !holds_something(&mut zip, part, kind)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
@@ -1033,39 +1013,13 @@ fn rels_source_dir(key: &str) -> Option<String> {
|
||||
|
||||
/// Whether a part has content: an element under its root, or for comments (whose authors list
|
||||
/// comes first) a comment, for VML a shape. Read only as far as needed; an unreadable part counts.
|
||||
/// `verdicts` remembers the answer per (part, named content); `budget` is the decompression this
|
||||
/// package's remaining checks may cause, and once it is spent the part counts as having content.
|
||||
fn holds_something(zip: &mut Zip<'_>, name: &str, kind: &str, verdicts: &mut HashMap<(String, &'static str), bool>, budget: &mut u64) -> bool {
|
||||
let wanted: Option<&'static [u8]> = match kind.to_ascii_lowercase().as_str() {
|
||||
"comments" => Some(b"comment"),
|
||||
"vmldrawing" => Some(b"shape"),
|
||||
_ => None,
|
||||
};
|
||||
let marker = match wanted {
|
||||
Some(b"comment") => "comment",
|
||||
Some(_) => "shape",
|
||||
None => "any",
|
||||
};
|
||||
if let Some(v) = verdicts.get(&(name.to_ascii_lowercase(), marker)) {
|
||||
return *v;
|
||||
}
|
||||
let v = holds_something_uncached(zip, name, marker, budget);
|
||||
verdicts.insert((name.to_ascii_lowercase(), marker), v);
|
||||
v
|
||||
}
|
||||
|
||||
fn holds_something_uncached(zip: &mut Zip<'_>, name: &str, marker: &'static str, budget: &mut u64) -> bool {
|
||||
fn holds_something(zip: &mut Zip<'_>, name: &str, kind: &str) -> bool {
|
||||
use quick_xml::events::Event;
|
||||
let Ok(f) = zip.by_name(name) else { return true };
|
||||
if *budget == 0 {
|
||||
return true;
|
||||
}
|
||||
let take = f.size().min(MAX_XML_PART).min(*budget);
|
||||
*budget -= take;
|
||||
let mut reader = quick_xml::Reader::from_reader(std::io::BufReader::new(f.take(take)));
|
||||
let wanted: Option<&'static [u8]> = match marker {
|
||||
"comment" => Some(b"comment"),
|
||||
"shape" => Some(b"shape"),
|
||||
let mut reader = quick_xml::Reader::from_reader(std::io::BufReader::new(f.take(MAX_XML_PART)));
|
||||
let wanted: Option<&[u8]> = match kind.to_ascii_lowercase().as_str() {
|
||||
"comments" => Some(b"comment"),
|
||||
"vmldrawing" => Some(b"shape"),
|
||||
_ => None,
|
||||
};
|
||||
let mut buf = Vec::new();
|
||||
@@ -1367,8 +1321,7 @@ mod tests {
|
||||
let path = std::env::temp_dir().join(format!("gridcraft-check-not-a-directory-{}", std::process::id()));
|
||||
std::fs::write(&path, "not a directory").unwrap();
|
||||
let mut discovery = Discovery::default();
|
||||
let mut seen = std::collections::HashSet::new();
|
||||
assert!(walk(&path, 0, &mut discovery, &mut seen));
|
||||
assert!(walk(&path, 0, &mut discovery));
|
||||
assert!(discovery.files.is_empty());
|
||||
assert_eq!(discovery.failures.len(), 1);
|
||||
assert_eq!(discovery.failures[0].path, path.to_string_lossy());
|
||||
|
||||
@@ -214,9 +214,7 @@ fn services(inbox: Inbox, ctx: egui::Context) -> Services {
|
||||
if let Some(w) = web_sys::window()
|
||||
&& let gridcraft_engine::links::LinkKind::External(url) = gridcraft_engine::links::classify(url)
|
||||
{
|
||||
// With `noopener` the new browsing context can't hold a reference back to the
|
||||
// editor window and navigate it while the user reads the linked page.
|
||||
let _ = w.open_with_url_and_target(&url, "_blank,noopener");
|
||||
let _ = w.open_with_url_and_target(&url, "_blank");
|
||||
}
|
||||
})),
|
||||
inbox: Some(inbox),
|
||||
|
||||
@@ -160,28 +160,21 @@ fn stroke_paths(pts: &[[f32; 2]], width: f32, dash: bool, closed: bool) -> Vec<P
|
||||
segs.push((*a, *b));
|
||||
}
|
||||
if dash {
|
||||
// The per-segment dash cap bounds each input line, but a caller-supplied vector of many
|
||||
// long dashed lines multiplies: together they would eagerly allocate megabytes of
|
||||
// geometry even when nothing of it reaches the canvas, so the total is capped too and
|
||||
// the line is drawn solid once the budget is gone.
|
||||
let (on, off) = (width * 4.0, width * 3.0);
|
||||
let mut out = Vec::new();
|
||||
let mut budget = 65_536usize;
|
||||
for (a, b) in segs {
|
||||
if budget == 0 {
|
||||
return vec![vec![pts.to_vec()]];
|
||||
}
|
||||
let (dx, dy) = (b[0] - a[0], b[1] - a[1]);
|
||||
let len = (dx * dx + dy * dy).sqrt();
|
||||
if !(len.is_finite() && len > 0.0) {
|
||||
continue;
|
||||
}
|
||||
let mut t = 0.0;
|
||||
while t < len && budget > 0 {
|
||||
let mut n = 0;
|
||||
while t < len && n < 10_000 {
|
||||
let e = (t + on).min(len);
|
||||
out.push(([a[0] + dx * t / len, a[1] + dy * t / len], [a[0] + dx * e / len, a[1] + dy * e / len]));
|
||||
t += on + off;
|
||||
budget -= 1;
|
||||
n += 1;
|
||||
}
|
||||
}
|
||||
segs = out;
|
||||
@@ -269,23 +262,3 @@ pub fn rasterize(prims: &[Prim], w: u32, h: u32, background: Rgba) -> Vec<u8> {
|
||||
}
|
||||
cv.buf
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
#[test]
|
||||
fn dashed_geometry_stays_within_the_dash_budget() {
|
||||
// Forty very long dashed lines, each allowed 10_000 dashes per segment: before the
|
||||
// shared cap they eagerly allocated ~1.6 MB of segment pairs × points and a like
|
||||
// number of quads before any of it reached a 4-pixel canvas (finding: dash expansion).
|
||||
let prims: Vec<Prim> =
|
||||
(0..40).map(|i| Prim::Line { pts: vec![[0.0, i as f32], [900.0, i as f32]], color: [0, 0, 0, 255], width: 1.0, dash: true }).collect();
|
||||
let img = rasterize(&prims, 2, 2, [255, 255, 255, 255]);
|
||||
assert_eq!(img.len(), 2 * 2 * 4);
|
||||
// And the geometry itself is bounded even for one huge dashed polyline.
|
||||
assert!(stroke_paths(&[[0.0, 0.0], [1e6, 0.0]], 1.0, true, false).len() <= 65_536);
|
||||
let prim = [Prim::Line { pts: vec![[0.0, 0.0], [10.0, 0.0]], color: [0, 0, 0, 255], width: 1.0, dash: true }];
|
||||
assert_eq!(rasterize(&prim, 1, 1, [0, 0, 0, 255]).len(), 4);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -350,21 +350,6 @@ mod tests {
|
||||
assert_eq!(split_sheet("A1"), (None, "A1"));
|
||||
}
|
||||
|
||||
/// #2312: device/backslash garbage in front of `!` is cosmetic at worst — `split_sheet` only
|
||||
/// splits at `!`, and every consumer matches the sheet part against workbook sheet names (it
|
||||
/// never reaches a filesystem path or a shell); the body half always stays a pure cell/range
|
||||
/// reference.
|
||||
#[test]
|
||||
fn split_sheet_leaves_device_names_inert_in_the_body() {
|
||||
for s in ["C:\\a\\file!A1", "CON!A1", "NUL!A1", "a[b]!A1"] {
|
||||
let (sheet, rest) = split_sheet(s);
|
||||
assert_eq!(CellRef::parse(rest), Some(CellRef::new(0, 0)), "{s:?}: body must stay a bare reference");
|
||||
let _ = sheet;
|
||||
}
|
||||
// A plain single-letter sheet is still split, not mistaken for a body.
|
||||
assert_eq!(split_sheet("C!A1").0, Some("C".to_string()));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn letters_roundtrip() {
|
||||
for (c, s) in [(0, "A"), (25, "Z"), (26, "AA"), (51, "AZ"), (52, "BA"), (701, "ZZ"), (702, "AAA"), (16383, "XFD")] {
|
||||
|
||||
@@ -371,16 +371,10 @@ fn parse_datetime(t: &str, sys: DateSystem) -> Option<(f64, Option<&'static str>
|
||||
if let Some((d, f)) = parse_date(t, sys) {
|
||||
return Some((d, Some(f)));
|
||||
}
|
||||
// date + time: split at the last space(s) before a time-looking tail. Only splits where the
|
||||
// tail could be a time are tried: a long tail can't parse, so this stays linear in the text
|
||||
// instead of reparsing every prefix for every space (one long text field quadratic).
|
||||
// date + time: split at the last space(s) before a time-looking tail.
|
||||
for (i, _) in t.match_indices(' ') {
|
||||
let Some(b) = t.get(i + 1..) else { continue };
|
||||
let bytes = b.as_bytes();
|
||||
if b.len() > 36 || !bytes.first().is_some_and(u8::is_ascii_digit) || !b.contains(':') {
|
||||
continue;
|
||||
}
|
||||
if let (Some((d, _)), Some((tf, _, _))) = (t.get(..i).and_then(|a| parse_date(a, sys)), parse_time(b)) {
|
||||
let (a, b) = (t.get(..i)?, t.get(i + 1..)?);
|
||||
if let (Some((d, _)), Some((tf, _, _))) = (parse_date(a, sys), parse_time(b)) {
|
||||
return Some((d + tf, Some("m/d/yyyy h:mm")));
|
||||
}
|
||||
}
|
||||
@@ -406,20 +400,6 @@ mod tests {
|
||||
assert_eq!(parse_number_text("1 1/2"), Some(1.5));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn overlong_time_fields_stay_text() {
|
||||
// #2168: all-digit field lengths are gated before allocation/derivation,
|
||||
// so hostile strings never spin the time-derivation machinery.
|
||||
let big = "9".repeat(500);
|
||||
for s in [format!("{big}:00"), format!("1:00:{big}"), format!("{big}:{big}:{big}"), format!("1:0{big}"), format!("00:00:{big}.1")] {
|
||||
assert_eq!(p(&s).value, Value::text(s.as_str()), "{s}");
|
||||
assert_eq!(parse_number_text(&s), None, "{s}");
|
||||
}
|
||||
// Within the gate, times still work.
|
||||
assert_eq!(p("01:02:03").value, Value::number(3.0 / 86_400.0 + 1.0 / 24.0 + 2.0 / 1440.0));
|
||||
assert_eq!(parse_number_text("13:14:15"), Some(0.5515625));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn non_digit_time_and_fraction_parts_stay_text() {
|
||||
for s in ["1:nan", "nan:30", "1:30:NaN", "1 nan/2", "1 1/nan", "inf:00", "1e1:00", "+5:00"] {
|
||||
|
||||
@@ -267,12 +267,6 @@ fn attr<'a>(attrs: &'a str, name: &str) -> Option<&'a str> {
|
||||
}
|
||||
}
|
||||
|
||||
/// Entities are at most `&` + 8 chars + `;` (a decimal code point at most 7 digits deep), so
|
||||
/// the semicolon is only looked for in the first 12 bytes — and `unwrap_or(rest)` never falls
|
||||
/// back to scanning the remainder of the text either (that turn re-scans per `&`, quadratic
|
||||
/// when a paste holds many bare `&`).
|
||||
const ENTITY_WINDOW: usize = 12;
|
||||
|
||||
/// HTML text to plain text: entities decoded, whitespace runs (and `<br>`) collapsed to one
|
||||
/// space, trimmed. The text only serves to match a cell with the tab-separated text.
|
||||
fn decode(raw: &str) -> String {
|
||||
@@ -281,9 +275,7 @@ fn decode(raw: &str) -> String {
|
||||
while let Some(amp) = rest.find('&') {
|
||||
out.push_str(rest.get(..amp).unwrap_or_default());
|
||||
rest = rest.get(amp..).unwrap_or_default();
|
||||
// `.min(len)` (not an `unwrap_or(rest)`) — a `get(..12)` that fails on a multibyte
|
||||
// boundary used to fall back to the complete suffix, re-scanning per '&'.
|
||||
let semi = rest.get(..ENTITY_WINDOW.min(rest.len())).and_then(|w| w.find(';'));
|
||||
let semi = rest.get(..12).unwrap_or(rest).find(';');
|
||||
let entity = semi.and_then(|e| rest.get(1..e));
|
||||
let ch = entity.and_then(|e| match e {
|
||||
"amp" => Some('&'),
|
||||
@@ -390,19 +382,6 @@ mod tests {
|
||||
assert_eq!(table_cells(quoted_gt, 1, 1), vec![vec![HtmlCell { text: "1500".into(), num: Some(1500.0) }]]);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn entities_recognized_only_within_the_twelve_byte_window() {
|
||||
// The window never scans the rest of the text for a semicolon: a bare '&' followed by
|
||||
// long plain text stays a bare '&', and one '&' after another is linear.
|
||||
assert_eq!(decode("a&lifier; b"), "a&lifier; b");
|
||||
assert_eq!(decode("&ampampampAMP;"), "&ampampampAMP;");
|
||||
// Inside the window the usual forms still decode, and a trailing bare '&' too.
|
||||
assert_eq!(decode("&"), "&");
|
||||
assert_eq!(decode("♪"), "\u{266A}");
|
||||
assert_eq!(decode("a&b c"), "a&b\u{a0}c");
|
||||
assert_eq!(decode("amp&"), "amp&");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn hostile_markup_does_not_panic() {
|
||||
for html in [
|
||||
|
||||
@@ -1034,11 +1034,6 @@ fn denied_script_step(id: &str, params: &Json) -> Option<String> {
|
||||
if id.starts_with("file.") || id.starts_with("link.") {
|
||||
return Some(format!("scripts can't run `{id}`: it reaches outside the workbook"));
|
||||
}
|
||||
// Scripts are scoped to their workbook: a step that switches the active window would
|
||||
// point every later step at another open document, so those are refused too.
|
||||
if id.starts_with("window.") {
|
||||
return Some(format!("scripts can't run `{id}`: it reaches outside the workbook"));
|
||||
}
|
||||
if params.get("path").is_some() {
|
||||
return Some(format!("scripts can't pass a file `path` to `{id}`"));
|
||||
}
|
||||
|
||||
@@ -233,23 +233,10 @@ fn with(p: &Json, key: &str, v: Json) -> Json {
|
||||
pub(crate) fn apply_style(s: &mut Session, p: &Json, f: impl Fn(&mut Style)) -> Result<Json> {
|
||||
let sheet = target_sheet(s, p)?;
|
||||
let ranges = target_ranges(s, p)?;
|
||||
let guarded = s.doc()?.wb.sheet(sheet).is_some_and(|sh| sh.protection.is_some());
|
||||
if s.doc()?.wb.sheet(sheet).is_some_and(|sh| sh.protection.as_ref().is_some_and(|pr| !pr.format_cells)) {
|
||||
return Err(EngineError::Other("The cell or chart you're trying to change is on a protected sheet.".into()));
|
||||
}
|
||||
edit(s, |cx| {
|
||||
// Formatting on a protected sheet keeps the cell's permission flags: a format call
|
||||
// that flips `locked` from the format permission would authorize content edits that
|
||||
// the password should still gate, so those flags may not change (Excel keeps that
|
||||
// checkbox disabled while the sheet is protected).
|
||||
const FLAG_GUARD: &str = "You can't change cell protection on a protected sheet. Unprotect it first.";
|
||||
macro_rules! guard_flags {
|
||||
($old:expr, $new:expr) => {
|
||||
if guarded && cx.wb.styles.get($old).protection != cx.wb.styles.get($new).protection {
|
||||
return Err(EngineError::Other(FLAG_GUARD.into()));
|
||||
}
|
||||
};
|
||||
}
|
||||
let mut cache: std::collections::HashMap<StyleId, StyleId> = std::collections::HashMap::new();
|
||||
for r in &ranges {
|
||||
let full_rows = r.is_full_rows();
|
||||
@@ -264,7 +251,6 @@ pub(crate) fn apply_style(s: &mut Session, p: &Json, f: impl Fn(&mut Style)) ->
|
||||
info.and_then(|i| i.style).unwrap_or_default()
|
||||
};
|
||||
let new = *cache.entry(cur).or_insert_with(|| cx.wb.styles.derive(cur, &f));
|
||||
guard_flags!(cur, new);
|
||||
let sh = cx.sheet_mut(sheet)?;
|
||||
let map = if full_rows && !full_cols { &mut sh.rows } else { &mut sh.cols };
|
||||
map.entry(line).or_default().style = Some(new);
|
||||
@@ -274,7 +260,6 @@ pub(crate) fn apply_style(s: &mut Session, p: &Json, f: impl Fn(&mut Style)) ->
|
||||
cx.wb.sheet(sheet).map(|sh| sh.cells.iter_range(*r).map(|(c, x)| (c, x.style)).collect()).unwrap_or_default();
|
||||
for (c, st) in &cells {
|
||||
let new = *cache.entry(*st).or_insert_with(|| cx.wb.styles.derive(*st, &f));
|
||||
guard_flags!(*st, new);
|
||||
cx.sheet_mut(sheet)?.set_style(*c, new);
|
||||
}
|
||||
// Formatting touched these cells: re-fit their rows (Excel auto-fits).
|
||||
@@ -291,7 +276,6 @@ pub(crate) fn apply_style(s: &mut Session, p: &Json, f: impl Fn(&mut Style)) ->
|
||||
let st = cx.wb.sheet(sheet).map(|sh| sh.style_id(c)).unwrap_or_default();
|
||||
let new = *cache.entry(st).or_insert_with(|| cx.wb.styles.derive(st, &f));
|
||||
if new != st {
|
||||
guard_flags!(st, new);
|
||||
cx.sheet_mut(sheet)?.set_style(c, new);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -552,15 +552,9 @@ pub fn auto_row_height(wb: &mut Workbook, si: usize, row: u32) {
|
||||
return;
|
||||
}
|
||||
let mut need: f32 = default;
|
||||
// The cell loop must not rescan every merge per cell: the merges touching this row are
|
||||
// picked out once, and per cell the candidates are re-checked (normally none or a couple;
|
||||
// a sheet with many merges and many saved columns would otherwise scan all of them per
|
||||
// cell, rows × merges × columns).
|
||||
let row_merges: Vec<gridcraft_core::RangeRef> =
|
||||
sh.merges.iter().copied().filter(|m| !m.is_single() && m.start.row <= row && row <= m.end.row).collect();
|
||||
for (c, cell) in sh.cells.row(row, 0, gridcraft_core::MAX_COLS - 1) {
|
||||
// Merged cells never auto-fit their rows, like Excel (column autofit skips them too).
|
||||
if row_merges.iter().any(|m| m.contains(CellRef::new(row, c))) {
|
||||
if sh.merge_at(CellRef::new(row, c)).is_some_and(|m| !m.is_single()) {
|
||||
continue;
|
||||
}
|
||||
let st = wb.styles.get(cell.style);
|
||||
|
||||
+90
-117
@@ -784,27 +784,22 @@ fn wrap_lines(text: &str, font: Font, size: f32, width: f32) -> Vec<String> {
|
||||
out.push(std::mem::take(&mut line));
|
||||
line = word.to_string();
|
||||
}
|
||||
// Break a single over-long word by characters. The cut that fits is found by binary
|
||||
// search (each try measures one candidate prefix), so an over-long word costs
|
||||
// O(n log n) measurements instead of re-measuring the whole cut per character for
|
||||
// every emitted line (`a` × 100k pasted text would be cubic worst case).
|
||||
// Break a single over-long word by characters.
|
||||
while gridcraft_pdf::text_width(font, size, &line) > width && line.chars().count() > 1 {
|
||||
let chars: Vec<char> = line.chars().collect();
|
||||
let (mut lo, mut hi) = (1usize, chars.len());
|
||||
while lo < hi {
|
||||
let mid = lo + (hi - lo).div_ceil(2);
|
||||
let cut: String = chars[..mid].iter().collect();
|
||||
if gridcraft_pdf::text_width(font, size, &cut) <= width {
|
||||
lo = mid;
|
||||
let mut cut = String::new();
|
||||
let mut rest = String::new();
|
||||
for ch in line.chars() {
|
||||
if rest.is_empty() && gridcraft_pdf::text_width(font, size, &format!("{cut}{ch}")) <= width || cut.is_empty() {
|
||||
cut.push(ch);
|
||||
} else {
|
||||
hi = mid - 1;
|
||||
rest.push(ch);
|
||||
}
|
||||
}
|
||||
if lo == chars.len() {
|
||||
if rest.is_empty() {
|
||||
break;
|
||||
}
|
||||
out.push(chars[..lo].iter().collect());
|
||||
line = chars[lo..].iter().collect();
|
||||
out.push(cut);
|
||||
line = rest;
|
||||
}
|
||||
if out.len() > 2000 {
|
||||
return out;
|
||||
@@ -1311,16 +1306,7 @@ fn decode_image(data: &[u8], max_w: u32, max_h: u32) -> Option<(Vec<u8>, u32, u3
|
||||
}
|
||||
|
||||
/// Draws charts, pictures, shapes and sparklines overlapping the body block of a page.
|
||||
/// `pics` memoizes decoded pictures for the whole render (one page at a time draws each
|
||||
/// picture again, but the decode happens once).
|
||||
fn draw_objects(
|
||||
pg: &mut Page,
|
||||
wb: &Workbook,
|
||||
si: usize,
|
||||
sh: &Sheet,
|
||||
b: &Block,
|
||||
pics: &mut std::collections::HashMap<(usize, usize), Option<(Vec<u8>, u32, u32)>>,
|
||||
) {
|
||||
fn draw_objects(pg: &mut Page, wb: &Workbook, si: usize, sh: &Sheet, b: &Block) {
|
||||
let block = b.range();
|
||||
let (sx0, sy0) = (sh.col_left(block.start.col), sh.row_top(block.start.row));
|
||||
let (sx1, sy1) = (sh.col_left(block.end.col.saturating_add(1)), sh.row_top(block.end.row.saturating_add(1)));
|
||||
@@ -1330,42 +1316,6 @@ fn draw_objects(
|
||||
let place = |r: (f64, f64, f64, f64)| (bx + ((r.0 - sx0) as f32) * b.k, by + ((r.1 - sy0) as f32) * b.k, r.2 as f32 * b.k, r.3 as f32 * b.k);
|
||||
pg.save();
|
||||
pg.clip_rect(bx, by, bw, bh);
|
||||
draw_sparklines(pg, wb, si, sh, b, block);
|
||||
for (imi, im) in sh.images.iter().enumerate() {
|
||||
let r = anchor_rect(sh, &im.anchor);
|
||||
if !visible(r) {
|
||||
continue;
|
||||
}
|
||||
let (x, y, w, h) = place(r);
|
||||
// About 150 dpi is plenty for print.
|
||||
let (mw, mh) = (((w / 72.0) * 150.0).ceil().clamp(1.0, 4000.0) as u32, ((h / 72.0) * 150.0).ceil().clamp(1.0, 4000.0) as u32);
|
||||
let decoded = pics.entry((si, imi)).or_insert_with(|| decode_image(&im.data, mw, mh));
|
||||
if let Some((px, pw, ph)) = decoded.as_ref() {
|
||||
pg.image_rgba(x, y, w, h, px, *pw, *ph);
|
||||
}
|
||||
}
|
||||
for shp in &sh.shapes {
|
||||
let r = anchor_rect(sh, &shp.anchor);
|
||||
if !visible(r) {
|
||||
continue;
|
||||
}
|
||||
let (x, y, w, h) = place(r);
|
||||
draw_shape(pg, wb, shp, b, (x, y, w, h));
|
||||
}
|
||||
for ch in &sh.charts {
|
||||
let r = anchor_rect(sh, &ch.anchor);
|
||||
if !visible(r) {
|
||||
continue;
|
||||
}
|
||||
let (x, y, _, _) = place(r);
|
||||
let data = gridcraft_chart::resolve(wb, si, ch);
|
||||
let prims = gridcraft_chart::render(ch, &data, r.2 as f32, r.3 as f32, &PdfMeasure);
|
||||
draw_prims(pg, &prims, Xf { ox: x, oy: y, k: b.k });
|
||||
}
|
||||
pg.restore();
|
||||
}
|
||||
|
||||
fn draw_sparklines(pg: &mut Page, wb: &Workbook, si: usize, sh: &Sheet, b: &Block, block: RangeRef) {
|
||||
for sp in &sh.sparklines {
|
||||
if !block.contains(sp.cell) {
|
||||
continue;
|
||||
@@ -1380,64 +1330,90 @@ fn draw_sparklines(pg: &mut Page, wb: &Workbook, si: usize, sh: &Sheet, b: &Bloc
|
||||
let prims = gridcraft_chart::render_sparkline(sp.kind, &vals, [color[0], color[1], color[2], 255], sp.markers, wpx, hpx);
|
||||
draw_prims(pg, &prims, Xf { ox: x, oy: y, k: b.k });
|
||||
}
|
||||
}
|
||||
|
||||
fn draw_shape(pg: &mut Page, wb: &Workbook, shp: &gridcraft_model::Shape, b: &Block, (x, y, w, h): (f32, f32, f32, f32)) {
|
||||
let fill = rgb_of(shp.fill, wb);
|
||||
let line = rgb_of(shp.line, wb).filter(|_| !shp.format.no_line).map(|c| (c, (0.75 * b.k).max(0.25)));
|
||||
match shp.kind {
|
||||
ShapeKind::Ellipse => {
|
||||
let pts: Vec<[f32; 2]> = (0..72)
|
||||
.map(|i| {
|
||||
let a = i as f32 / 72.0 * std::f32::consts::TAU;
|
||||
[x + w / 2.0 + w / 2.0 * a.cos(), y + h / 2.0 + h / 2.0 * a.sin()]
|
||||
})
|
||||
.collect();
|
||||
pg.polygon(&pts, fill, line);
|
||||
for im in &sh.images {
|
||||
let r = anchor_rect(sh, &im.anchor);
|
||||
if !visible(r) {
|
||||
continue;
|
||||
}
|
||||
ShapeKind::Triangle => pg.polygon(&[[x + w / 2.0, y], [x + w, y + h], [x, y + h]], fill, line),
|
||||
ShapeKind::Line | ShapeKind::Arrow => {
|
||||
let c = line.map(|l| l.0).or(fill).unwrap_or([0, 0, 0]);
|
||||
pg.stroke_line(x, y, x + w, y + h, (1.0 * b.k).max(0.3), c);
|
||||
if shp.kind == ShapeKind::Arrow {
|
||||
let ang = h.atan2(w);
|
||||
let al = 8.0 * b.k;
|
||||
let p = |da: f32| [x + w - al * (ang + da).cos(), y + h - al * (ang + da).sin()];
|
||||
pg.polygon(&[[x + w, y + h], p(0.4), p(-0.4)], Some(c), None);
|
||||
let (x, y, w, h) = place(r);
|
||||
// About 150 dpi is plenty for print.
|
||||
let (mw, mh) = (((w / 72.0) * 150.0).ceil().clamp(1.0, 4000.0) as u32, ((h / 72.0) * 150.0).ceil().clamp(1.0, 4000.0) as u32);
|
||||
if let Some((px, pw, ph)) = decode_image(&im.data, mw, mh) {
|
||||
pg.image_rgba(x, y, w, h, &px, pw, ph);
|
||||
}
|
||||
}
|
||||
for shp in &sh.shapes {
|
||||
let r = anchor_rect(sh, &shp.anchor);
|
||||
if !visible(r) {
|
||||
continue;
|
||||
}
|
||||
let (x, y, w, h) = place(r);
|
||||
let fill = rgb_of(shp.fill, wb);
|
||||
let line = rgb_of(shp.line, wb).filter(|_| !shp.format.no_line).map(|c| (c, (0.75 * b.k).max(0.25)));
|
||||
match shp.kind {
|
||||
ShapeKind::Ellipse => {
|
||||
let pts: Vec<[f32; 2]> = (0..72)
|
||||
.map(|i| {
|
||||
let a = i as f32 / 72.0 * std::f32::consts::TAU;
|
||||
[x + w / 2.0 + w / 2.0 * a.cos(), y + h / 2.0 + h / 2.0 * a.sin()]
|
||||
})
|
||||
.collect();
|
||||
pg.polygon(&pts, fill, line);
|
||||
}
|
||||
ShapeKind::Triangle => pg.polygon(&[[x + w / 2.0, y], [x + w, y + h], [x, y + h]], fill, line),
|
||||
ShapeKind::Line | ShapeKind::Arrow => {
|
||||
let c = line.map(|l| l.0).or(fill).unwrap_or([0, 0, 0]);
|
||||
pg.stroke_line(x, y, x + w, y + h, (1.0 * b.k).max(0.3), c);
|
||||
if shp.kind == ShapeKind::Arrow {
|
||||
let ang = h.atan2(w);
|
||||
let al = 8.0 * b.k;
|
||||
let p = |da: f32| [x + w - al * (ang + da).cos(), y + h - al * (ang + da).sin()];
|
||||
pg.polygon(&[[x + w, y + h], p(0.4), p(-0.4)], Some(c), None);
|
||||
}
|
||||
}
|
||||
_ => {
|
||||
pg.polygon(&[[x, y], [x + w, y], [x + w, y + h], [x, y + h]], fill, line);
|
||||
}
|
||||
}
|
||||
_ => {
|
||||
pg.polygon(&[[x, y], [x + w, y], [x + w, y + h], [x, y + h]], fill, line);
|
||||
if !shp.text.is_empty() {
|
||||
use gridcraft_model::{ShapeTextAlign, ShapeTextAnchor};
|
||||
let fm = &shp.format;
|
||||
let text_box = shp.kind == ShapeKind::TextBox;
|
||||
let size = fm.font_size.unwrap_or(11.0).clamp(1.0, 4000.0) * b.scale;
|
||||
let font = Font::styled(Font::family_for(fm.font.as_deref().unwrap_or("")), fm.bold, fm.italic);
|
||||
let color = fm.text_color.and_then(|c| rgb_of(c, wb)).unwrap_or([0, 0, 0]);
|
||||
let align = fm.align.unwrap_or(if text_box { ShapeTextAlign::Left } else { ShapeTextAlign::Center });
|
||||
let anchor = fm.anchor.unwrap_or(if text_box { ShapeTextAnchor::Top } else { ShapeTextAnchor::Middle });
|
||||
let lines = wrap_lines(&shp.text, font, size, (w - 4.0 * b.k).max(size));
|
||||
let block_h = lines.len().min(500) as f32 * size * 1.2;
|
||||
let mut yy = match anchor {
|
||||
ShapeTextAnchor::Top => y + 2.0 * b.k,
|
||||
ShapeTextAnchor::Middle => y + (h - block_h) / 2.0,
|
||||
ShapeTextAnchor::Bottom => y + h - 2.0 * b.k - block_h,
|
||||
} + size * 0.8;
|
||||
for l in lines.iter().take(500) {
|
||||
let lw = gridcraft_pdf::text_width(font, size, l);
|
||||
let lx = match align {
|
||||
ShapeTextAlign::Left => x + 2.0 * b.k,
|
||||
ShapeTextAlign::Center => x + (w - lw) / 2.0,
|
||||
ShapeTextAlign::Right => x + w - 2.0 * b.k - lw,
|
||||
};
|
||||
pg.text(lx, yy, size, font, color, l);
|
||||
yy += size * 1.2;
|
||||
}
|
||||
}
|
||||
}
|
||||
if shp.text.is_empty() {
|
||||
return;
|
||||
}
|
||||
use gridcraft_model::{ShapeTextAlign, ShapeTextAnchor};
|
||||
let fm = &shp.format;
|
||||
let text_box = shp.kind == ShapeKind::TextBox;
|
||||
let size = fm.font_size.unwrap_or(11.0).clamp(1.0, 4000.0) * b.scale;
|
||||
let font = Font::styled(Font::family_for(fm.font.as_deref().unwrap_or("")), fm.bold, fm.italic);
|
||||
let color = fm.text_color.and_then(|c| rgb_of(c, wb)).unwrap_or([0, 0, 0]);
|
||||
let align = fm.align.unwrap_or(if text_box { ShapeTextAlign::Left } else { ShapeTextAlign::Center });
|
||||
let anchor = fm.anchor.unwrap_or(if text_box { ShapeTextAnchor::Top } else { ShapeTextAnchor::Middle });
|
||||
let lines = wrap_lines(&shp.text, font, size, (w - 4.0 * b.k).max(size));
|
||||
let block_h = lines.len().min(500) as f32 * size * 1.2;
|
||||
let mut yy = match anchor {
|
||||
ShapeTextAnchor::Top => y + 2.0 * b.k,
|
||||
ShapeTextAnchor::Middle => y + (h - block_h) / 2.0,
|
||||
ShapeTextAnchor::Bottom => y + h - 2.0 * b.k - block_h,
|
||||
} + size * 0.8;
|
||||
for l in lines.iter().take(500) {
|
||||
let lw = gridcraft_pdf::text_width(font, size, l);
|
||||
let lx = match align {
|
||||
ShapeTextAlign::Left => x + 2.0 * b.k,
|
||||
ShapeTextAlign::Center => x + (w - lw) / 2.0,
|
||||
ShapeTextAlign::Right => x + w - 2.0 * b.k - lw,
|
||||
};
|
||||
pg.text(lx, yy, size, font, color, l);
|
||||
yy += size * 1.2;
|
||||
for ch in &sh.charts {
|
||||
let r = anchor_rect(sh, &ch.anchor);
|
||||
if !visible(r) {
|
||||
continue;
|
||||
}
|
||||
let (x, y, _, _) = place(r);
|
||||
let data = gridcraft_chart::resolve(wb, si, ch);
|
||||
let prims = gridcraft_chart::render(ch, &data, r.2 as f32, r.3 as f32, &PdfMeasure);
|
||||
draw_prims(pg, &prims, Xf { ox: x, oy: y, k: b.k });
|
||||
}
|
||||
pg.restore();
|
||||
}
|
||||
|
||||
fn draw_headings(pg: &mut Page, b: &Block, col_heads: bool, row_heads: bool, at: (f32, f32), size: (f32, f32)) {
|
||||
@@ -1552,9 +1528,6 @@ pub fn render_pdf(wb: &Workbook, layouts: &[SheetLayout], file: &str, path: &str
|
||||
let mut page_no = 0usize;
|
||||
let mut pending_links = Vec::new();
|
||||
let mut printed_blocks = Vec::new();
|
||||
// Decoded pictures are reused across the pages their anchor covers: a floating picture
|
||||
// spanning 40 pages was decoded (and pushed into the PDF) 40 times (#2389).
|
||||
let mut pics: std::collections::HashMap<(usize, usize), Option<(Vec<u8>, u32, u32)>> = std::collections::HashMap::new();
|
||||
for lay in layouts {
|
||||
let Some(sh) = wb.sheet(lay.sheet) else { continue };
|
||||
let mut cf = if sh.cond_formats.is_empty() { None } else { Some(crate::cf::CfCache::new()) };
|
||||
@@ -1591,7 +1564,7 @@ pub fn render_pdf(wb: &Workbook, layouts: &[SheetLayout], file: &str, path: &str
|
||||
y += seg_size(sh, *rs, true) * k;
|
||||
}
|
||||
if let Some(b) = &body {
|
||||
draw_objects(pg, wb, lay.sheet, sh, b, &mut pics);
|
||||
draw_objects(pg, wb, lay.sheet, sh, b);
|
||||
}
|
||||
if lay.settings.headings {
|
||||
let fill = [0xF2, 0xF2, 0xF2];
|
||||
|
||||
@@ -401,10 +401,8 @@ fn move_columns(s: &mut Session, p: &Json) -> Result<Json> {
|
||||
{
|
||||
return Err(EngineError::Other("To prevent possible loss of data, columns cannot be shifted off of the worksheet.".into()));
|
||||
}
|
||||
// The checks must authorize the sheet this command modifies (`target_sheet`), not whatever
|
||||
// sheet is active: both carry the command's own `sheet` parameter on.
|
||||
check_lines_protection(s, &json!({"cols": RangeRef::cols(before, before).a1(), "sheet": p.get("sheet")}), Axis::Cols, before, count, true)?;
|
||||
check_lines_protection(s, &json!({"cols": RangeRef::cols(from, end - 1).a1(), "sheet": p.get("sheet")}), Axis::Cols, from, count, false)?;
|
||||
check_lines_protection(s, &json!({"cols": RangeRef::cols(before, before).a1()}), Axis::Cols, before, count, true)?;
|
||||
check_lines_protection(s, &json!({"cols": RangeRef::cols(from, end - 1).a1()}), Axis::Cols, from, count, false)?;
|
||||
|
||||
let moved_from = if before <= from { from + count } else { from };
|
||||
if moved_from.checked_add(count).is_none_or(|end| end > MAX_COLS) {
|
||||
|
||||
+4
-126
@@ -448,11 +448,6 @@ impl std::fmt::Write for ClipboardHtml {
|
||||
/// Most rows Save as Web Page writes. A sheet whose used range is taller fails instead of losing cells.
|
||||
const HTML_EXPORT_MAX_ROWS: u32 = 100_001;
|
||||
|
||||
/// Rows × columns of one web-page table. Save as Web Page writes every kept row's full used
|
||||
/// column span into one growing String, so a sparse sheet with a lone far-right cell (or a
|
||||
/// tall, wide one) would otherwise be gigabytes of `<td></td>` before finishing (#2387).
|
||||
const HTML_EXPORT_MAX_CELLS: u64 = 4_000_000;
|
||||
|
||||
/// A sheet as a standalone HTML table (Save as Web Page). A sheet with pictures gets the sheet's
|
||||
/// own column widths and row heights so each picture and chart (inline SVG) can sit at its anchor.
|
||||
pub fn to_html(wb: &Workbook, sheet: usize) -> Result<String> {
|
||||
@@ -484,17 +479,6 @@ pub fn to_html(wb: &Workbook, sheet: usize) -> Result<String> {
|
||||
r.height()
|
||||
)));
|
||||
}
|
||||
// The table is one growing String: every kept row writes a `<td>` for the whole used
|
||||
// column span (unless the sheet has no row that is that wide). A tall, wide sheet
|
||||
// (or a sparse one with a lone far-right cell) would be gigabytes before finishing,
|
||||
// so the rectangle itself is bounded (#2387).
|
||||
if u64::from(r.height()) * u64::from(r.width()) > HTML_EXPORT_MAX_CELLS {
|
||||
return Err(EngineError::Other(format!(
|
||||
"the sheet is too tall × wide for a web page ({} × {} cells, at most {HTML_EXPORT_MAX_CELLS}). Save as .xlsx or CSV instead.",
|
||||
r.height(),
|
||||
r.width()
|
||||
)));
|
||||
}
|
||||
if fixed {
|
||||
out.push_str("<colgroup>");
|
||||
for col in (r.start.col..=r.end.col).filter(|&c| !sh.is_col_hidden(c)) {
|
||||
@@ -604,27 +588,14 @@ fn script_sandbox_check(path: &str) -> Result<()> {
|
||||
Ok(())
|
||||
}
|
||||
|
||||
/// The largest native file the reader still touches: its own cap plus one, used to stop
|
||||
/// reads whose bytes arrive faster than metadata can claim (device files, growing files).
|
||||
const MAX_READ_BYTES: u64 = 2 * 1024 * 1024 * 1024;
|
||||
|
||||
#[cfg(not(target_arch = "wasm32"))]
|
||||
pub fn read_file(path: &str) -> Result<Vec<u8>> {
|
||||
use std::io::Read;
|
||||
script_sandbox_check(path)?;
|
||||
let f = std::fs::File::open(path).map_err(|e| EngineError::Other(format!("{path}: {e}")))?;
|
||||
// A symlink to an endless device reports length 0 but streams forever, so metadata
|
||||
// alone never bounds the read: require a real file and cap the bytes actually taken.
|
||||
if !f.metadata().map_err(|e| EngineError::Other(format!("{path}: {e}")))?.is_file() {
|
||||
return Err(EngineError::Other(format!("{path}: only regular files can be opened")));
|
||||
}
|
||||
let mut buf = Vec::new();
|
||||
let mut limited = (&f).take(MAX_READ_BYTES + 1);
|
||||
limited.read_to_end(&mut buf).map_err(|e| EngineError::Other(format!("{path}: {e}")))?;
|
||||
if buf.len() as u64 > MAX_READ_BYTES {
|
||||
let meta = std::fs::metadata(path).map_err(|e| EngineError::Other(format!("{path}: {e}")))?;
|
||||
if meta.len() > 2 * 1024 * 1024 * 1024 {
|
||||
return Err(EngineError::Other("files larger than 2 GB aren't supported".into()));
|
||||
}
|
||||
Ok(buf)
|
||||
std::fs::read(path).map_err(|e| EngineError::Other(format!("{path}: {e}")))
|
||||
}
|
||||
|
||||
#[cfg(target_arch = "wasm32")]
|
||||
@@ -635,35 +606,10 @@ pub fn read_file(path: &str) -> Result<Vec<u8>> {
|
||||
|
||||
#[cfg(not(target_arch = "wasm32"))]
|
||||
pub fn write_file(path: &str, bytes: &[u8]) -> Result<()> {
|
||||
use std::io::Write;
|
||||
script_sandbox_check(path)?;
|
||||
// Write to a temporary file, then rename: a crash never leaves a half-written workbook.
|
||||
// The temporary file is created exclusively, so an existing file or symlink at that
|
||||
// path (planted by another local user) makes the save fail instead of truncating the
|
||||
// link's target; and it inherits the destination's access restrictions, so saving a
|
||||
// private workbook doesn't widen who can read it.
|
||||
let tmp = format!("{path}.sctmp");
|
||||
let prior = std::fs::metadata(path).ok().map(|m| m.permissions());
|
||||
let mut opts = std::fs::OpenOptions::new();
|
||||
opts.write(true).create_new(true);
|
||||
#[cfg(unix)]
|
||||
if let Some(p) = &prior {
|
||||
use std::os::unix::fs::{OpenOptionsExt, PermissionsExt};
|
||||
opts.mode(p.mode());
|
||||
}
|
||||
let mut f = opts.open(&tmp).map_err(|e| {
|
||||
let _ = std::fs::remove_file(&tmp);
|
||||
EngineError::Other(format!("{path}: {e}"))
|
||||
})?;
|
||||
#[cfg(not(unix))]
|
||||
if let Some(p) = &prior {
|
||||
let _ = f.set_permissions(p.clone());
|
||||
}
|
||||
f.write_all(bytes).and_then(|()| f.flush()).map_err(|e| {
|
||||
let _ = std::fs::remove_file(&tmp);
|
||||
EngineError::Other(format!("{path}: {e}"))
|
||||
})?;
|
||||
std::fs::rename(&tmp, path).map_err(|e| {
|
||||
std::fs::write(&tmp, bytes).and_then(|()| std::fs::rename(&tmp, path)).map_err(|e| {
|
||||
// A write or rename that fails (full disk, the workbook locked by another app) leaves
|
||||
// no temporary file beside the user's workbook.
|
||||
let _ = std::fs::remove_file(&tmp);
|
||||
@@ -765,74 +711,6 @@ mod tests {
|
||||
range_html(wb, 0, r, &crate::display::range_text(wb, 0, r)).unwrap()
|
||||
}
|
||||
|
||||
#[cfg(not(target_arch = "wasm32"))]
|
||||
#[test]
|
||||
fn save_fails_upfront_when_the_temp_name_is_taken() {
|
||||
// Another local user who can write the directory must not be able to point the
|
||||
// save's predictable temporary name at their own file: creation is exclusive.
|
||||
let dir = std::env::temp_dir().join("gridcraft-fs-safety");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
std::fs::create_dir_all(&dir).unwrap();
|
||||
let victim = dir.join("victim.txt");
|
||||
std::fs::write(&victim, b"original").unwrap();
|
||||
let dest = dir.join("book.xlsx");
|
||||
std::os::unix::fs::symlink(&victim, dir.join("book.xlsx.sctmp")).unwrap();
|
||||
let err = write_file(dest.to_str().unwrap(), b"new").unwrap_err();
|
||||
assert_eq!(std::fs::read(&victim).unwrap(), b"original");
|
||||
assert!(!dir.join("book.xlsx").exists(), "{err}");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
#[cfg(not(target_arch = "wasm32"))]
|
||||
#[cfg(unix)]
|
||||
#[test]
|
||||
fn save_keeps_the_destination_access_restrictions() {
|
||||
use std::os::unix::fs::PermissionsExt;
|
||||
let dir = std::env::temp_dir().join("gridcraft-fs-safety");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
std::fs::create_dir_all(&dir).unwrap();
|
||||
let dest = dir.join("private.xlsx");
|
||||
std::fs::write(&dest, b"v1").unwrap();
|
||||
std::fs::set_permissions(&dest, std::fs::Permissions::from_mode(0o600)).unwrap();
|
||||
write_file(dest.to_str().unwrap(), b"v2").unwrap();
|
||||
let mode = std::fs::metadata(&dest).unwrap().permissions().mode() & 0o777;
|
||||
assert_eq!(mode, 0o600);
|
||||
// And a brand-new file is created with the restrictive, not permissive, default.
|
||||
let fresh = dir.join("new.xlsx");
|
||||
write_file(fresh.to_str().unwrap(), b"v").unwrap();
|
||||
let mode = std::fs::metadata(&fresh).unwrap().permissions().mode() & 0o777;
|
||||
assert!(mode & 0o022 == 0, "mode {mode:o}");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
#[cfg(not(target_arch = "wasm32"))]
|
||||
#[test]
|
||||
fn open_refuses_directories_and_unknowable_lengths() {
|
||||
let dir = std::env::temp_dir().join("gridcraft-fs-safety");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
std::fs::create_dir_all(&dir).unwrap();
|
||||
assert!(read_file(dir.to_str().unwrap()).is_err(), "a directory is not a regular file");
|
||||
assert!(read_file(dir.join("missing").to_str().unwrap()).is_err());
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn html_export_refuses_an_outsized_grid_instead_of_growing_forever() {
|
||||
// #2387: the used rectangle alone bounds the output (every row writes the whole
|
||||
// column span), so a sparse sheet with one cell at A1 and another far right
|
||||
// 300 rows down is 4.9M cells → refused before a byte of `<td>` is written.
|
||||
let mut wb = Workbook::new();
|
||||
let sh = wb.sheet_mut(0).unwrap();
|
||||
sh.set_value(CellRef::new(0, 0), gridcraft_core::Value::Text("start".into()));
|
||||
sh.set_value(CellRef::new(300, 16383), gridcraft_core::Value::Text("corner".into()));
|
||||
let err = to_html(&wb, 0).unwrap_err();
|
||||
assert!(err.to_string().contains("too tall × wide"), "{err}");
|
||||
// A normal sheet is fine.
|
||||
let mut ok = Workbook::new();
|
||||
ok.sheet_mut(0).unwrap().set_value(CellRef::new(5, 2), gridcraft_core::Value::Text("t".into()));
|
||||
assert!(to_html(&ok, 0).is_ok());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn html_export_embeds_images_as_data_uris() {
|
||||
let mut wb = Workbook::new();
|
||||
|
||||
@@ -4604,60 +4604,4 @@ fn html_export_without_pictures_or_charts_has_no_fixed_geometry() {
|
||||
let html = crate::io::to_html(&s.doc().unwrap().wb, 0).unwrap();
|
||||
assert!(!html.contains("<colgroup>") && !html.contains("table-layout") && !html.contains("<svg") && !html.contains("<img"), "{html}");
|
||||
}
|
||||
#[test]
|
||||
fn format_cells_permission_on_a_protected_sheet_cannot_unlock_cells() {
|
||||
// Formatting may be allowed on a protected sheet; the permission can't serve as a key to
|
||||
// content edits, so the format cells may not drop a cell's `locked` flag.
|
||||
let mut s = s();
|
||||
s.execute("cell.set", json!({"cell": "B2", "input": "hi"})).unwrap();
|
||||
s.execute("review.protectSheet", json!({"formatCells": true})).unwrap();
|
||||
s.execute("home.fontName", json!({"name": "Arial"})).unwrap();
|
||||
let err = s.execute("home.lockCell", json!({"range": "B2", "on": false})).expect_err("unlock through formatting");
|
||||
assert!(err.to_string().contains("protected"), "{err}");
|
||||
assert!(s.execute("cell.set", json!({"cell": "B2", "input": "x"})).is_err(), "B2 stays locked");
|
||||
// The same calls run once the sheet is unprotected.
|
||||
s.execute("review.unprotectSheet", json!({})).unwrap();
|
||||
s.execute("home.lockCell", json!({"range": "B2", "on": false})).unwrap();
|
||||
s.execute("cell.set", json!({"cell": "B2", "input": "x"})).unwrap();
|
||||
assert_eq!(v(&s, "B2"), Value::text("x"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn move_columns_checks_the_protection_of_the_named_sheet() {
|
||||
// The move's protection checks must authorize the sheet it modifies, not whichever sheet
|
||||
// happens to be active while the command names another one.
|
||||
let mut s = s();
|
||||
s.execute("home.insertSheet", json!({"name": "P"})).unwrap();
|
||||
s.execute("range.setValues", json!({"range": "A1", "values": [[1, 2, 3, 4]]})).unwrap();
|
||||
s.execute("review.protectSheet", json!({})).unwrap();
|
||||
s.execute("sheet.activate", json!({"sheet": 0})).unwrap();
|
||||
let err = s.execute("sheet.moveColumns", json!({"sheet": "P", "cols": "B:C", "before": 4})).expect_err("moving columns on a protected sheet");
|
||||
assert!(err.to_string().contains("protected"), "{err}");
|
||||
let pv: Vec<Value> = {
|
||||
let d = s.doc().unwrap();
|
||||
let sh = &d.wb.sheets[d.wb.sheet_index("P").unwrap()];
|
||||
(0..4).map(|c| sh.value(CellRef::new(0, c as u32))).collect()
|
||||
};
|
||||
assert_eq!(pv, vec![Value::number(1.0), Value::number(2.0), Value::number(3.0), Value::number(4.0)]);
|
||||
// The active, unprotected sheet moves fine.
|
||||
s.execute("range.setValues", json!({"range": "A1", "values": [["A", "B", "C", "D"]]})).unwrap();
|
||||
s.execute("sheet.moveColumns", json!({"cols": "B:B", "before": 4})).unwrap();
|
||||
assert_eq!(v(&s, "A1"), Value::text("A"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_script_cannot_switch_windows() {
|
||||
// Stored scripts are scoped to their workbook: a step that activates another window would
|
||||
// point every later step at another open document, so it is refused before any step runs.
|
||||
let mut s = s();
|
||||
s.execute("automate.newScript", json!({"name": "t", "commands": [{"command": "window.activate", "params": {"index": 0}}]})).unwrap();
|
||||
s.execute("automate.trustScripts", json!({"trusted": true})).unwrap();
|
||||
let err = s.execute("automate.runScript", json!({"name": "t"})).expect_err("script switching windows");
|
||||
assert!(err.to_string().contains("window.activate"), "{err}");
|
||||
// A benign script runs.
|
||||
s.execute("automate.newScript", json!({"name": "ok", "commands": [{"command": "cell.set", "params": {"cell": "A1", "input": "=1+1"}}]})).unwrap();
|
||||
s.execute("automate.runScript", json!({"name": "ok"})).unwrap();
|
||||
assert_eq!(v(&s, "A1"), Value::number(2.0));
|
||||
}
|
||||
|
||||
mod seams;
|
||||
|
||||
@@ -95,13 +95,6 @@ impl Cache {
|
||||
}
|
||||
None => (None, u32::MAX),
|
||||
};
|
||||
// Eviction only frees entries used before this frame, so a workbook covered in same-frame
|
||||
// pictures can leave the cache unable to make room. Then draw the texture without
|
||||
// retaining it: the handle is dropped at the end of the frame and the GPU memory goes
|
||||
// with it, instead of the budget being outflanked (finding: cache admission).
|
||||
if self.entries.len() >= MAX_CACHE_ITEMS || self.bytes.saturating_add(bytes) > MAX_CACHE_BYTES {
|
||||
return texture;
|
||||
}
|
||||
self.bytes = self.bytes.saturating_add(bytes);
|
||||
self.entries.insert(key, Entry { texture: texture.clone(), bytes, side, used: frame });
|
||||
texture
|
||||
@@ -219,24 +212,13 @@ mod tests {
|
||||
}
|
||||
assert!(cache.bytes <= MAX_CACHE_BYTES, "{} bytes kept", cache.bytes);
|
||||
assert!(cache.entries.len() < data.len());
|
||||
// Many on screen at once, over the budget: each still gets a texture this frame, but the
|
||||
// cache cannot grow past its budget whatever the workbook shows. Eight of the 16 MB
|
||||
// textures fit the 128 MB budget, and those stay pinned; the rest re-upload each frame
|
||||
// instead of being retained over it (finding: cache admission).
|
||||
// Two frames pass so the rehearsal's entries are old enough to be evictable.
|
||||
frame(&ctx);
|
||||
frame(&ctx);
|
||||
// Many on screen at once, over the budget: every one keeps its texture, frame after frame.
|
||||
let shown: Vec<_> = data.iter().enumerate().take(12).map(|(i, d)| cache.texture(&ctx, key(1, i as u32, d), d, 2048).unwrap().id()).collect();
|
||||
for _ in 0..3 {
|
||||
frame(&ctx);
|
||||
for (i, d) in data.iter().enumerate().take(12) {
|
||||
let id = cache.texture(&ctx, key(1, i as u32, d), d, 2048).unwrap().id();
|
||||
if i < 8 {
|
||||
assert_eq!(id, shown[i], "a picture that fits keeps its texture");
|
||||
}
|
||||
assert_eq!(cache.texture(&ctx, key(1, i as u32, d), d, 2048).map(|t| t.id()), Some(shown[i]));
|
||||
}
|
||||
assert!(cache.bytes <= MAX_CACHE_BYTES, "{} bytes kept", cache.bytes);
|
||||
assert!(cache.entries.len() <= 8, "{} entries kept", cache.entries.len());
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -26,9 +26,6 @@ const BLOCK: usize = 16;
|
||||
pub(crate) const BLOCK_VERIFIER_INPUT: [u8; 8] = [0xFE, 0xA7, 0xD2, 0x76, 0x3B, 0x4B, 0x9E, 0x79];
|
||||
pub(crate) const BLOCK_VERIFIER_HASH: [u8; 8] = [0xD7, 0xAA, 0x0F, 0x6D, 0x30, 0x61, 0x34, 0x4E];
|
||||
pub(crate) const BLOCK_KEY: [u8; 8] = [0x14, 0x6E, 0x0B, 0xE7, 0xAB, 0xAC, 0xD0, 0xD6];
|
||||
/// The two dataIntegrity block keys ([MS-OFFCRYPTO] 2.3.4.14), one per HMAC field.
|
||||
pub(crate) const BLOCK_HMAC_KEY: [u8; 8] = [0x5F, 0xB2, 0xAD, 0x01, 0x0C, 0xB9, 0xE1, 0xF6];
|
||||
pub(crate) const BLOCK_HMAC_VALUE: [u8; 8] = [0xA0, 0x67, 0x7F, 0x02, 0xB2, 0x2C, 0x84, 0x33];
|
||||
|
||||
/// The password key encryptor's URI (the other kind is certificate-based).
|
||||
pub(crate) const PASSWORD_ENCRYPTOR: &str = "http://schemas.microsoft.com/office/2006/keyEncryptor/password";
|
||||
@@ -168,14 +165,6 @@ impl Hash {
|
||||
Hash::Sha512 => run::<sha2::Sha512>(parts),
|
||||
}
|
||||
}
|
||||
|
||||
/// The bytes the algorithm's compression function takes: 128 for SHA-384/512, else 64.
|
||||
fn block_size(self) -> usize {
|
||||
match self {
|
||||
Hash::Sha384 | Hash::Sha512 => 128,
|
||||
Hash::Sha1 | Hash::Sha256 => 64,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// H0 = H(salt ‖ UTF-16LE(password)), then Hn = H(LE32(n - 1) ‖ Hn-1) for `spin` rounds.
|
||||
@@ -195,19 +184,6 @@ pub(crate) fn derive(hash: Hash, a: &[u8], b: &[u8], len: usize) -> Vec<u8> {
|
||||
k
|
||||
}
|
||||
|
||||
/// HMAC-SHA-`hash` ([RFC2104]), as Agile's dataIntegrity uses.
|
||||
pub(crate) fn hmac(hash: Hash, key: &[u8], message: &[u8]) -> Vec<u8> {
|
||||
let bs = hash.block_size();
|
||||
let mut k = key.to_vec();
|
||||
if k.len() > bs {
|
||||
k = hash.digest(&[&k]);
|
||||
}
|
||||
k.resize(bs, 0);
|
||||
let pad = |x: u8| k.iter().map(|b| b ^ x).collect::<Vec<u8>>();
|
||||
let inner = hash.digest(&[&pad(0x36), message]);
|
||||
hash.digest(&[&pad(0x5C), &inner])
|
||||
}
|
||||
|
||||
/// The Standard encryption key ([MS-OFFCRYPTO] 2.3.4.7): SHA-1 iterated 50,000 times, the
|
||||
/// block-0 hash, then the 0x36 / 0x5C expansion, truncated to `key_len`.
|
||||
pub(crate) fn standard_key(password: &str, salt: &[u8], key_len: usize) -> Vec<u8> {
|
||||
@@ -392,9 +368,8 @@ impl Params {
|
||||
}
|
||||
}
|
||||
|
||||
/// What an Agile descriptor holds: the package's cipher, the password key encryptor with its
|
||||
/// three encrypted values (verifier input, verifier hash, package key), and the package's
|
||||
/// integrity pair, which Excel always writes and the reader requires.
|
||||
/// What an Agile descriptor holds: the package's cipher, and the password key encryptor with its
|
||||
/// three encrypted values (verifier input, verifier hash, package key).
|
||||
struct AgileDescriptor {
|
||||
key_data: Params,
|
||||
key: Params,
|
||||
@@ -402,8 +377,6 @@ struct AgileDescriptor {
|
||||
verifier_input: Vec<u8>,
|
||||
verifier_hash: Vec<u8>,
|
||||
key_value: Vec<u8>,
|
||||
hmac_key: Vec<u8>,
|
||||
hmac_value: Vec<u8>,
|
||||
}
|
||||
|
||||
/// `xml` is the XML encryption descriptor after the 8-byte version and reserved field.
|
||||
@@ -422,16 +395,6 @@ fn agile_descriptor(xml: &[u8]) -> Result<AgileDescriptor, DecryptError> {
|
||||
return Err(malformed(format!("a spin count of {spin}, more than the {MAX_SPIN_COUNT} allowed")));
|
||||
}
|
||||
let field = |name: &str| base64(encryptor.attr(name).ok_or_else(|| malformed(format!("no {name}")))?);
|
||||
// A descriptor without dataIntegrity can't prove its ciphertext is the one the password
|
||||
// protects: the verifier only authenticates the key, not the package ([MS-OFFCRYPTO] 2.3.4.14).
|
||||
let integrity = root.child("dataIntegrity").ok_or_else(|| malformed("no dataIntegrity"))?;
|
||||
let hmac_field = |attr: &str| {
|
||||
let v = base64(integrity.attr(attr).ok_or_else(|| malformed(format!("no {attr}")))?)?;
|
||||
if v.is_empty() || v.len() % BLOCK != 0 {
|
||||
return Err(malformed(format!("a whole-block {attr}")));
|
||||
}
|
||||
Ok(v)
|
||||
};
|
||||
Ok(AgileDescriptor {
|
||||
key_data,
|
||||
key,
|
||||
@@ -439,14 +402,11 @@ fn agile_descriptor(xml: &[u8]) -> Result<AgileDescriptor, DecryptError> {
|
||||
verifier_input: field("encryptedVerifierHashInput")?,
|
||||
verifier_hash: field("encryptedVerifierHashValue")?,
|
||||
key_value: field("encryptedKeyValue")?,
|
||||
hmac_key: hmac_field("encryptedHmacKey")?,
|
||||
hmac_value: hmac_field("encryptedHmacValue")?,
|
||||
})
|
||||
}
|
||||
|
||||
fn agile(xml: &[u8], package: &[u8], password: &str) -> Result<Vec<u8>, DecryptError> {
|
||||
let descriptor = agile_descriptor(xml)?;
|
||||
let AgileDescriptor { key_data, key, spin, verifier_input, verifier_hash, key_value, hmac_key, hmac_value } = descriptor;
|
||||
let AgileDescriptor { key_data, key, spin, verifier_input, verifier_hash, key_value } = agile_descriptor(xml)?;
|
||||
let h = iterated_hash(key.hash, password, &key.salt, spin);
|
||||
let iv = key.iv(None);
|
||||
let field = |data: &[u8], block_key: &[u8]| -> Result<Vec<u8>, DecryptError> {
|
||||
@@ -463,18 +423,6 @@ fn agile(xml: &[u8], package: &[u8], password: &str) -> Result<Vec<u8>, DecryptE
|
||||
}
|
||||
let secret = field(&key_value, &BLOCK_KEY)?;
|
||||
let aes = Aes::new(secret.get(..key_data.key_len).ok_or_else(|| malformed("a short encryptedKeyValue"))?)?;
|
||||
// Package integrity, [MS-OFFCRYPTO] 2.3.4.14: both HMAC fields are encrypted with the
|
||||
// package key (the verifier above only authenticates the password), keyed by a random
|
||||
// salt-size array, computed over the whole EncryptedPackage stream, StreamSize field included.
|
||||
let mut mac_field = hmac_key;
|
||||
aes.decrypt_cbc(&key_data.iv(Some(&BLOCK_HMAC_KEY)), &mut mac_field)?;
|
||||
mac_field.truncate(key_data.salt.len());
|
||||
let mut declared = hmac_value;
|
||||
aes.decrypt_cbc(&key_data.iv(Some(&BLOCK_HMAC_VALUE)), &mut declared)?;
|
||||
let mac = hmac(key_data.hash, &mac_field, package);
|
||||
if declared.get(..mac.len()) != Some(mac.as_slice()) {
|
||||
return Err(malformed("the decrypted data failed its integrity check"));
|
||||
}
|
||||
let (size, data) = package_parts(package)?;
|
||||
let mut out = Vec::with_capacity(size);
|
||||
for (i, segment) in data.chunks(SEGMENT).enumerate() {
|
||||
|
||||
@@ -6,9 +6,6 @@ use super::*;
|
||||
use crate::testing::{agile_info, encrypt_agile, encrypt_standard, with_agile_header};
|
||||
|
||||
const PASSWORD: &str = "Pass1234";
|
||||
/// 16 zero bytes (whole AES blocks), a stand-in for a real integrity field where the values
|
||||
/// can't be reached anyway (scheme errors, descriptor-parse damage).
|
||||
const HMAC_STUB: &str = "AAAAAAAAAAAAAAAAAAAAAA==";
|
||||
|
||||
fn hex(s: &str) -> Vec<u8> {
|
||||
(0..s.len()).step_by(2).map(|i| u8::from_str_radix(&s[i..i + 2], 16).unwrap()).collect()
|
||||
@@ -110,9 +107,9 @@ fn unsupported_schemes_are_named() {
|
||||
assert_eq!(unsupported(&rc4), "RC4");
|
||||
assert_eq!(unsupported(&[3, 0, 3, 0, 0, 0, 0, 0]), "Extensible");
|
||||
assert_eq!(unsupported(&[1, 0, 1, 0, 0, 0, 0, 0]), "version 1.1");
|
||||
let des = agile_info(PASSWORD, 1000, HMAC_STUB, HMAC_STUB).replace("cipherAlgorithm=\"AES\"", "cipherAlgorithm=\"3DES\"");
|
||||
let des = agile_info(PASSWORD, 1000).replace("cipherAlgorithm=\"AES\"", "cipherAlgorithm=\"3DES\"");
|
||||
assert_eq!(unsupported(&with_agile_header(&des)), "3DES");
|
||||
let cert = agile_info(PASSWORD, 1000, HMAC_STUB, HMAC_STUB).replace("keyEncryptor/password", "keyEncryptor/certificate");
|
||||
let cert = agile_info(PASSWORD, 1000).replace("keyEncryptor/password", "keyEncryptor/certificate");
|
||||
assert_eq!(unsupported(&with_agile_header(&cert)), "certificate");
|
||||
}
|
||||
|
||||
@@ -139,7 +136,7 @@ fn damaged_files_are_malformed_not_panics() {
|
||||
ragged[..8].copy_from_slice(&len.to_le_bytes());
|
||||
assert!(is_malformed(decrypt_streams(&info, &ragged, PASSWORD)));
|
||||
}
|
||||
let xml = agile_info(PASSWORD, 1000, HMAC_STUB, HMAC_STUB);
|
||||
let xml = agile_info(PASSWORD, 1000);
|
||||
let bad = |s: String| is_malformed(decrypt_streams(&with_agile_header(&s), &[0; 8], PASSWORD));
|
||||
assert!(bad(xml.replace("spinCount=\"1000\"", "spinCount=\"4000000000\"")));
|
||||
assert!(bad(xml.replace("saltValue=\"", "saltValue=\"*")));
|
||||
@@ -158,76 +155,3 @@ fn non_ascii_passwords_hash_as_utf16le_like_msoffcrypto() {
|
||||
hex("2bca96ca0a7e8d88d9314a2eb1819a1ece94ff9ce34266ff82f891a88c4240c4c9f39c729b7c5c410c27fc8e2b7ad90d8c005eb68a603377b8a4ab858041c76a")
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn hmac_matches_the_rfc2104_test_vectors() {
|
||||
use crate::crypto::hmac as mac;
|
||||
let h = |v: Vec<u8>| v.iter().map(|b| format!("{b:02x}")).collect::<String>();
|
||||
// RFC 2202 test case 1 (HMAC-SHA-1) and HMAC-SHA-256 vectors.
|
||||
assert_eq!(h(mac(Hash::Sha1, &[0x0b; 20], b"Hi There")), "b617318655057264e28bc0b6fb378c8ef146be00");
|
||||
assert_eq!(h(mac(Hash::Sha256, &[0x0b; 20], b"Hi There")), "b0344c61d8db38535ca8afceaf0bf12b881dc200c9833da726e9376c2e32cff7");
|
||||
// RFC 4231 test case 5 (HMAC-SHA-256): a key longer than the block size is hashed first.
|
||||
assert_eq!(
|
||||
h(mac(Hash::Sha256, &[0xaa; 131], b"Test Using Larger Than Block-Size Key - Hash Key First")),
|
||||
"60e431591ee0b67f0d8a26aacbf5b77f8e0bc6213728c5140546040f0ee37f54"
|
||||
);
|
||||
}
|
||||
|
||||
/// Strips the `<dataIntegrity …/>` element from an Agile descriptor.
|
||||
fn without_integrity(info: &[u8]) -> Vec<u8> {
|
||||
let xml = std::str::from_utf8(&info[8..]).unwrap().to_string();
|
||||
let start = xml.find("<dataIntegrity").unwrap();
|
||||
let end = start + xml[start..].find("/>").unwrap() + 2;
|
||||
let stripped = format!("{}{}", &xml[..start], &xml[end..]);
|
||||
with_agile_header(&stripped)
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn integrity_is_required_and_enforced() {
|
||||
let plain = payload(300);
|
||||
let (info, package) = encrypt_agile(&plain, PASSWORD, 1000);
|
||||
// An attacker can't just delete the integrity element to get the package verified.
|
||||
assert!(matches!(decrypt_streams(&without_integrity(&info), &package, PASSWORD),
|
||||
Err(DecryptError::Malformed(m)) if m.contains("dataIntegrity")));
|
||||
// An empty or ragged field is damage, not a pass.
|
||||
for bad in [("encryptedHmacKey", ""), ("encryptedHmacKey", "AA=="), ("encryptedHmacValue", "")] {
|
||||
let xml =
|
||||
agile_info(PASSWORD, 1000, HMAC_STUB, HMAC_STUB).replace(&format!("{}=\"{HMAC_STUB}\"", bad.0), &format!("{}=\"{}\"", bad.0, bad.1));
|
||||
let r = decrypt_streams(&with_agile_header(&xml), &package, PASSWORD);
|
||||
assert!(matches!(r, Err(DecryptError::Malformed(_))), "{}=\"{}\"", bad.0, bad.1);
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn modified_ciphertext_fails_the_integrity_check() {
|
||||
let plain = payload(300);
|
||||
let (info, package) = encrypt_agile(&plain, PASSWORD, 1000);
|
||||
// The correct password proves the verifier can't vouch for a modified package.
|
||||
for at in [8usize, 100, package.len() - 5] {
|
||||
let mut tampered = package.clone();
|
||||
tampered[at] ^= 0x80;
|
||||
assert!(
|
||||
matches!(decrypt_streams(&info, &tampered, PASSWORD),
|
||||
Err(DecryptError::Malformed(m)) if m.contains("integrity")),
|
||||
"byte {at}"
|
||||
);
|
||||
}
|
||||
// And neither can the descriptor: a swapped integrity value is refused.
|
||||
let xml = String::from_utf8_lossy(&info[8..]).to_string();
|
||||
let at = xml.find("encryptedHmacValue=\"").unwrap() + "encryptedHmacValue=\"".len();
|
||||
let mut xml: Vec<u8> = xml.into_bytes();
|
||||
xml[at] ^= b'A' ^ b'B';
|
||||
assert!(matches!(decrypt_streams(&with_agile_header(std::str::from_utf8(&xml).unwrap()), &package, PASSWORD), Err(DecryptError::Malformed(_))));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn integrity_matches_the_stream_not_the_plaintext() {
|
||||
// The MAC covers the whole EncryptedPackage stream, StreamSize field included: changing the
|
||||
// stated size alone must fail, while the actual bytes round-trip.
|
||||
let plain = payload(4200);
|
||||
let (info, package) = encrypt_agile(&plain, PASSWORD, 1000);
|
||||
let mut ours = package.clone();
|
||||
ours[..8].copy_from_slice(&((plain.len() + 1) as u64).to_le_bytes());
|
||||
assert!(matches!(decrypt_streams(&info, &ours, PASSWORD), Err(DecryptError::Malformed(m)) if m.contains("integrity")));
|
||||
assert_eq!(decrypt_streams(&info, &package, PASSWORD).unwrap(), plain);
|
||||
}
|
||||
|
||||
@@ -74,24 +74,14 @@ impl<'a> Package<'a> {
|
||||
let remaining = MAX_TOTAL.saturating_sub(self.total);
|
||||
let limit = maximum.min(remaining);
|
||||
let mut out = Vec::with_capacity(f.size().min(16 * 1024 * 1024) as usize);
|
||||
let read = (&mut f).take(limit + 1).read_to_end(&mut out).map_err(|e| {
|
||||
// A read that fails partway has already spent decompression work and buffered data
|
||||
// the package budget never sees: count the allowance as spent, so a loop over many
|
||||
// broken parts can't decompress it over and over.
|
||||
self.total = MAX_TOTAL;
|
||||
IoError::Zip(format!("{part}: {e}"))
|
||||
});
|
||||
(&mut f).take(limit + 1).read_to_end(&mut out).map_err(|e| IoError::Zip(format!("{part}: {e}")))?;
|
||||
if out.len() as u64 > limit {
|
||||
// The allowance was spent getting here; make the exhaustion terminal instead of
|
||||
// letting later reads pay the same decompression cost again.
|
||||
self.total = MAX_TOTAL;
|
||||
return Err(IoError::TooLarge(if remaining < maximum {
|
||||
"package expands to more than 1 GB".into()
|
||||
} else {
|
||||
format!("{part} expands to more than {maximum} bytes")
|
||||
}));
|
||||
}
|
||||
read?;
|
||||
self.total += out.len() as u64;
|
||||
Ok(Some(out))
|
||||
}
|
||||
@@ -211,20 +201,4 @@ mod tests {
|
||||
assert_eq!(split_dir("xl/workbook.xml"), ("xl", "workbook.xml"));
|
||||
assert_eq!(split_dir("x.xml"), ("", "x.xml"));
|
||||
}
|
||||
|
||||
/// An over-budget read must spend the package's allowance once and for all: loops over many
|
||||
/// broken parts (a sheet's optional VML relationships, for one) otherwise re-decompress the
|
||||
/// remaining allowance again and again.
|
||||
#[test]
|
||||
fn a_failed_part_ends_the_package_budget() {
|
||||
let z = crate::tests::make_zip(&[("a.xml", b"0123456789"), ("b.xml", b"hello")]);
|
||||
let mut pkg = Package::open(&z).unwrap();
|
||||
assert!(pkg.read("a.xml").unwrap().is_some());
|
||||
// Nothing left to spend within the per-part limit either.
|
||||
pkg.total = MAX_TOTAL - 1;
|
||||
assert!(pkg.read("b.xml").is_err());
|
||||
assert_eq!(pkg.total, MAX_TOTAL, "the failed read is charged");
|
||||
// Oversized repeated attempts stay cheap errors now: no allowance is left to pay for.
|
||||
assert!(matches!(pkg.read("b.xml"), Err(IoError::TooLarge(_))));
|
||||
}
|
||||
}
|
||||
|
||||
@@ -33,62 +33,12 @@ struct RowState {
|
||||
/// Dynamic-array spill ranges and legacy array ranges: cached values inside them (other than
|
||||
/// the anchor) are not stored, so the formula can fill them again.
|
||||
dynamic: Vec<RangeRef>,
|
||||
/// Cells inside the ranges above, for a cap before a pathological sheet proves it itself.
|
||||
dynamic_cells: u64,
|
||||
/// Row-membership checks served so far, for the work cap above.
|
||||
dynamic_checks: u64,
|
||||
dropped: u64,
|
||||
rows: u64,
|
||||
}
|
||||
|
||||
/// Whether `pos` lies in a known dynamic range that isn't its anchor: the cell's cached value
|
||||
/// gives way to the formula that will fill it. Charges every scanned range against one work
|
||||
/// budget, then gives up consulting the ranges at all: the worst kind of file may then import
|
||||
/// a few cached spill values, but it never stalls the whole import.
|
||||
fn in_dynamic(st: &mut RowState, pos: CellRef) -> bool {
|
||||
if st.dynamic_checks > MAX_DYNAMIC_CHECKS || st.dynamic.is_empty() {
|
||||
return false;
|
||||
}
|
||||
let ranges = st.dynamic.len();
|
||||
let hit = st.dynamic.iter().any(|r| r.contains(pos) && r.start != pos);
|
||||
st.dynamic_checks = st.dynamic_checks.saturating_add(ranges as u64);
|
||||
hit
|
||||
}
|
||||
|
||||
/// Records a dynamic-array range, so its cells outside the anchor import as empty and the
|
||||
/// formula fills them again. Repeated identical anchors (a damaged or hostile sheet may repeat
|
||||
/// the same record) are recorded once, and count and area are capped: every later cell inside
|
||||
/// known range territory pays a scan proportional to what is recorded here.
|
||||
fn push_dynamic(st: &mut RowState, range: RangeRef) {
|
||||
if st.dynamic.contains(&range) {
|
||||
return;
|
||||
}
|
||||
let cells = range.count();
|
||||
if st.dynamic.len() >= MAX_DYNAMIC_RANGES || st.dynamic_cells.saturating_add(cells) > MAX_DYNAMIC_CELLS {
|
||||
return;
|
||||
}
|
||||
st.dynamic_cells = st.dynamic_cells.saturating_add(cells);
|
||||
st.dynamic.push(range);
|
||||
}
|
||||
|
||||
const MAX_HYPERLINK_CELLS: u64 = 10_000;
|
||||
/// Target and tooltip bytes kept for hyperlinks of one sheet, over all cells: the same URL is
|
||||
/// copied for every cell of its range, so length × cells must stay bounded whatever the file.
|
||||
const MAX_HYPERLINK_BYTES: u64 = 8 * 1024 * 1024;
|
||||
const MAX_COLUMN_ASSIGNMENTS: u64 = 1_000_000;
|
||||
/// Most dynamic-array and legacy-array ranges of one sheet.
|
||||
const MAX_DYNAMIC_RANGES: usize = 65_536;
|
||||
/// Worst-case cells inside the dynamic ranges the sheet declares. Beyond it further ranges are
|
||||
/// dropped: they would cost their storage in every membership lookup that follows.
|
||||
const MAX_DYNAMIC_CELLS: u64 = 16_000_000;
|
||||
/// Most row-membership checks the sheet's ranges may serve. Beyond it the remaining cells are
|
||||
/// imported with their cached values instead of being blanked for a spill that may still come,
|
||||
/// so adversarial range counts can't make each cell scan a long list (a real workbook has a
|
||||
/// handful of ranges).
|
||||
const MAX_DYNAMIC_CHECKS: u64 = 16_000_000;
|
||||
/// Most conditional-format range cells one sheet keeps, over all rules of an element: every
|
||||
/// rule copies its element's whole range list, so rules × ranges must stay bounded.
|
||||
const MAX_COND_CELLS: u64 = 4_000_000;
|
||||
|
||||
pub fn read_sheet(cx: &mut Ctx<'_>, part: &str, name: &str) -> Result<Sheet, IoError> {
|
||||
let bytes = cx.pkg.read(part)?.ok_or_else(|| IoError::Format(format!("missing part {part}")))?;
|
||||
@@ -511,7 +461,7 @@ fn read_cell(cx: &mut Ctx<'_>, sheet: &mut Sheet, st: &mut RowState, c: &El, pos
|
||||
}
|
||||
let dynamic = c.attr("cm").is_some();
|
||||
let formula = c.child("f").and_then(|f| read_formula(cx, st, f, pos, dynamic));
|
||||
let value = if formula.is_none() && in_dynamic(st, pos) { Value::Empty } else { value };
|
||||
let value = if formula.is_none() && st.dynamic.iter().any(|r| r.contains(pos) && r.start != pos) { Value::Empty } else { value };
|
||||
sheet.cells.set(pos, Cell { value, formula: formula.map(Arc::new), style });
|
||||
}
|
||||
|
||||
@@ -552,7 +502,7 @@ fn read_formula_text(cx: &mut Ctx<'_>, st: &mut RowState, f: &El, pos: CellRef,
|
||||
let mut fm = Formula::new(text);
|
||||
let range = f.attr("ref").and_then(RangeRef::parse).unwrap_or(RangeRef::cell(pos));
|
||||
if !range.is_single() {
|
||||
push_dynamic(st, range);
|
||||
st.dynamic.push(range);
|
||||
}
|
||||
if !dynamic {
|
||||
fm.array = Some(range);
|
||||
@@ -722,12 +672,6 @@ fn read_cf(cx: &mut Ctx<'_>, e: &El, sheet: &mut Sheet) {
|
||||
if let Some(id) = ext_id {
|
||||
cx.cf_ext_ids.push((id, sheet.cond_formats.len()));
|
||||
}
|
||||
let rules = sheet.cond_formats.len();
|
||||
let area = ranges.iter().map(RangeRef::count).sum::<u64>().saturating_mul(1 + rules as u64);
|
||||
if area > MAX_COND_CELLS {
|
||||
cx.warn(format!("conditional formats past {MAX_COND_CELLS} kept range cells were dropped"));
|
||||
break;
|
||||
}
|
||||
sheet.cond_formats.push(CondFormat {
|
||||
ranges: ranges.clone(),
|
||||
rule,
|
||||
@@ -790,7 +734,6 @@ fn read_validation(v: &El) -> Option<Validation> {
|
||||
|
||||
fn read_hyperlinks(cx: &mut Ctx<'_>, name: &str, e: &El, rels: &[Rel], sheet: &mut Sheet) {
|
||||
let mut cells = 0u64;
|
||||
let mut bytes = 0u64;
|
||||
for h in e.kids("hyperlink") {
|
||||
let Some(range) = h.attr("ref").and_then(RangeRef::parse) else { continue };
|
||||
let mut target = h.attr("id").and_then(|id| rels.iter().find(|r| r.id == id)).map(|r| r.target.clone()).unwrap_or_default();
|
||||
@@ -806,19 +749,13 @@ fn read_hyperlinks(cx: &mut Ctx<'_>, name: &str, e: &El, rels: &[Rel], sheet: &m
|
||||
if target.is_empty() {
|
||||
continue;
|
||||
}
|
||||
let link = Hyperlink { target: target.clone(), tooltip: h.attr("tooltip").map(str::to_string), external };
|
||||
let link_bytes = (target.len() + link.tooltip.as_ref().map_or(0, |t| t.len())) as u64;
|
||||
let link = Hyperlink { target, tooltip: h.attr("tooltip").map(str::to_string), external };
|
||||
for c in range.iter() {
|
||||
if cells >= MAX_HYPERLINK_CELLS {
|
||||
cx.warn(format!("sheet \"{name}\": hyperlinks past the first {MAX_HYPERLINK_CELLS} cells were dropped"));
|
||||
return;
|
||||
}
|
||||
if bytes.saturating_add(link_bytes) > MAX_HYPERLINK_BYTES {
|
||||
cx.warn(format!("sheet \"{name}\": hyperlink targets past {MAX_HYPERLINK_BYTES} bytes were dropped"));
|
||||
return;
|
||||
}
|
||||
cells += 1;
|
||||
bytes += link_bytes;
|
||||
sheet.hyperlinks.insert(c, link.clone());
|
||||
}
|
||||
}
|
||||
|
||||
@@ -9,8 +9,8 @@
|
||||
use aes::cipher::{BlockCipherEncrypt, KeyInit};
|
||||
|
||||
use crate::crypto::{
|
||||
BLOCK_HMAC_KEY, BLOCK_HMAC_VALUE, BLOCK_KEY, BLOCK_VERIFIER_HASH, BLOCK_VERIFIER_INPUT, F_AES, F_CRYPTOAPI, Hash, PASSWORD_ENCRYPTOR, Params,
|
||||
SEGMENT, derive, hmac, iterated_hash, standard_key,
|
||||
BLOCK_KEY, BLOCK_VERIFIER_HASH, BLOCK_VERIFIER_INPUT, F_AES, F_CRYPTOAPI, Hash, PASSWORD_ENCRYPTOR, Params, SEGMENT, derive, iterated_hash,
|
||||
standard_key,
|
||||
};
|
||||
|
||||
/// The spin count `encrypt_workbook` uses: Excel's default.
|
||||
@@ -21,7 +21,6 @@ const KEY_DATA_SALT: [u8; 16] = [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27,
|
||||
const SECRET_KEY: [u8; 32] =
|
||||
[64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95];
|
||||
const VERIFIER: [u8; 16] = [32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47];
|
||||
const HMAC_SALT: [u8; 16] = [112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127];
|
||||
|
||||
/// `package` (normally an `.xlsx` zip) as Excel stores it with a password: Agile encryption
|
||||
/// (AES-256, SHA-512) inside a compound file.
|
||||
@@ -203,9 +202,8 @@ fn package(size: usize, body: Vec<u8>) -> Vec<u8> {
|
||||
p
|
||||
}
|
||||
|
||||
/// The Agile XML descriptor: AES-256, SHA-512, a password encryptor salted with 0..16, and the
|
||||
/// dataIntegrity pair (`hmac_key_b64`/`hmac_value_b64`) a compliant file carries.
|
||||
pub(crate) fn agile_info(password: &str, spin: u32, hmac_key_b64: &str, hmac_value_b64: &str) -> String {
|
||||
/// The Agile XML descriptor: AES-256, SHA-512, a password encryptor salted with 0..16.
|
||||
pub(crate) fn agile_info(password: &str, spin: u32) -> String {
|
||||
let key = Params { hash: Hash::Sha512, salt: PASSWORD_SALT.to_vec(), key_len: 32 };
|
||||
let h = iterated_hash(Hash::Sha512, password, &key.salt, spin);
|
||||
let iv = key.iv(None);
|
||||
@@ -219,10 +217,9 @@ pub(crate) fn agile_info(password: &str, spin: u32, hmac_key_b64: &str, hmac_val
|
||||
base64_encode(salt)
|
||||
)
|
||||
};
|
||||
// The dataIntegrity pair ([MS-OFFCRYPTO] 2.3.4.14), encrypted with the package key.
|
||||
format!(
|
||||
"<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"yes\"?>\r\n<encryption xmlns=\"http://schemas.microsoft.com/office/2006/encryption\" xmlns:p=\"http://schemas.microsoft.com/office/2006/keyEncryptor/password\"><keyData {}/><dataIntegrity encryptedHmacKey=\"{hmac_key_b64}\" encryptedHmacValue=\"{hmac_value_b64}\"/><keyEncryptors><keyEncryptor uri=\"{PASSWORD_ENCRYPTOR}\"><p:encryptedKey spinCount=\"{spin}\" {} encryptedVerifierHashInput=\"{input}\" encryptedVerifierHashValue=\"{value}\" encryptedKeyValue=\"{secret}\"/></keyEncryptor></keyEncryptors></encryption>",
|
||||
params(KEY_DATA_SALT.as_slice()),
|
||||
"<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"yes\"?>\r\n<encryption xmlns=\"http://schemas.microsoft.com/office/2006/encryption\" xmlns:p=\"http://schemas.microsoft.com/office/2006/keyEncryptor/password\"><keyData {}/><keyEncryptors><keyEncryptor uri=\"{PASSWORD_ENCRYPTOR}\"><p:encryptedKey spinCount=\"{spin}\" {} encryptedVerifierHashInput=\"{input}\" encryptedVerifierHashValue=\"{value}\" encryptedKeyValue=\"{secret}\"/></keyEncryptor></keyEncryptors></encryption>",
|
||||
params(&KEY_DATA_SALT),
|
||||
params(&PASSWORD_SALT),
|
||||
)
|
||||
}
|
||||
@@ -234,20 +231,14 @@ pub(crate) fn with_agile_header(xml: &str) -> Vec<u8> {
|
||||
info
|
||||
}
|
||||
|
||||
/// `EncryptionInfo` and `EncryptedPackage` for `plain`, Agile (AES-256, SHA-512), with the
|
||||
/// dataIntegrity HMAC over the whole EncryptedPackage stream as Excel writes it.
|
||||
/// `EncryptionInfo` and `EncryptedPackage` for `plain`, Agile (AES-256, SHA-512).
|
||||
pub(crate) fn encrypt_agile(plain: &[u8], password: &str, spin: u32) -> (Vec<u8>, Vec<u8>) {
|
||||
let key_data = Params { hash: Hash::Sha512, salt: KEY_DATA_SALT.to_vec(), key_len: 32 };
|
||||
let mut body = vec![];
|
||||
for (i, segment) in plain.chunks(SEGMENT).enumerate() {
|
||||
body.extend(encrypt_cbc(&SECRET_KEY, &key_data.iv(Some(&(i as u32).to_le_bytes())), segment));
|
||||
}
|
||||
let stream = package(plain.len(), body);
|
||||
let (hmac_key_b64, hmac_value_b64) = (
|
||||
base64_encode(&encrypt_cbc(&SECRET_KEY, &key_data.iv(Some(&BLOCK_HMAC_KEY)), &HMAC_SALT)),
|
||||
base64_encode(&encrypt_cbc(&SECRET_KEY, &key_data.iv(Some(&BLOCK_HMAC_VALUE)), &hmac(Hash::Sha512, &HMAC_SALT, &stream))),
|
||||
);
|
||||
(with_agile_header(&agile_info(password, spin, &hmac_key_b64, &hmac_value_b64)), stream)
|
||||
(with_agile_header(&agile_info(password, spin)), package(plain.len(), body))
|
||||
}
|
||||
|
||||
/// `EncryptionInfo` and `EncryptedPackage` for `plain`, Standard (AES-128, SHA-1).
|
||||
|
||||
@@ -56,8 +56,7 @@ fn schemes_no_password_opens_are_named_before_asking_for_one() {
|
||||
let file = |info: &[u8]| compound_file(&[("EncryptionInfo", info), ("EncryptedPackage", &[0; 8])]);
|
||||
assert_eq!(check_encryption(&file(&[3, 0, 2, 0, 4, 0, 0, 0])), unsupported("RC4"));
|
||||
assert_eq!(check_encryption(&file(&[4, 0, 3, 0, 0, 0, 0, 0])), unsupported("Extensible"));
|
||||
let cert =
|
||||
agile_info("pw", 1000, "AAAAAAAAAAAAAAAAAAAAAA==", "AAAAAAAAAAAAAAAAAAAAAA==").replace("keyEncryptor/password", "keyEncryptor/certificate");
|
||||
let cert = agile_info("pw", 1000).replace("keyEncryptor/password", "keyEncryptor/certificate");
|
||||
assert_eq!(check_encryption(&file(&with_agile_header(&cert))), unsupported("certificate"));
|
||||
assert!(matches!(check_encryption(&file(&[4, 0])), Err(DecryptError::Malformed(_))));
|
||||
// Rights management (IRM) keeps an EncryptedPackage but no EncryptionInfo.
|
||||
|
||||
+22
-61
@@ -13,7 +13,7 @@ mod formula;
|
||||
mod ftab;
|
||||
mod records;
|
||||
|
||||
use std::collections::{HashMap, HashSet};
|
||||
use std::collections::HashMap;
|
||||
use std::sync::Arc;
|
||||
|
||||
use gridcraft_core::{CellRef, DateSystem, MAX_COLS, MAX_ROWS, RangeRef, Value};
|
||||
@@ -59,14 +59,6 @@ pub(crate) fn is_xls(bytes: &[u8]) -> bool {
|
||||
cf.has_root_stream("Workbook") || cf.has_root_stream("Book")
|
||||
}
|
||||
|
||||
/// Most shared formulas one sheet expands, and the longest template text one attach may
|
||||
/// produce: a template shared by its followers was costing every one of them a full decode and
|
||||
/// print however often that could never succeed, so each anchor is tried once and failed ones
|
||||
/// are skipped for the rest of the sheet. Beyond the cap the remaining references keep their
|
||||
/// cached values.
|
||||
const MAX_SHARED_ATTACHES: usize = 262_144;
|
||||
const MAX_SHARED_TEMPLATE: usize = 8192;
|
||||
|
||||
pub(crate) struct Budget {
|
||||
cells: usize,
|
||||
text: usize,
|
||||
@@ -486,56 +478,6 @@ fn rk(v: u32) -> Value {
|
||||
Value::number(if v & 1 != 0 { n / 100.0 } else { n })
|
||||
}
|
||||
|
||||
/// Attaches every pending formula to its anchor's template, returning how many shared
|
||||
/// templates were expanded. An anchor whose template can't be imported is refused once:
|
||||
/// its followers re-decode the same template for nothing, so they are skipped instead (they
|
||||
/// keep their cached values).
|
||||
fn resolve_pending(
|
||||
st: &mut SheetState<'_>,
|
||||
arrays: &HashMap<CellRef, (RangeRef, Parsed)>,
|
||||
shared: &HashMap<CellRef, Parsed>,
|
||||
pending: &[Pending],
|
||||
) -> usize {
|
||||
let mut failed: HashSet<CellRef> = HashSet::new();
|
||||
let mut shared_attaches = 0usize;
|
||||
for p in pending {
|
||||
if let Some((range, f)) = arrays.get(&p.anchor) {
|
||||
if p.pos == p.anchor {
|
||||
st.attach(p.pos, f, Mode::Cell, Some(*range));
|
||||
}
|
||||
} else if let Some(f) = shared.get(&p.anchor) {
|
||||
if failed.contains(&p.anchor) || shared_attaches >= MAX_SHARED_ATTACHES {
|
||||
continue;
|
||||
}
|
||||
shared_attaches += 1;
|
||||
st.attach(p.pos, f, Mode::Shared, None);
|
||||
let imported = st.sheet.cell(p.pos).and_then(|c| c.formula.as_ref()).is_some_and(|f| f.text.chars().count() <= MAX_SHARED_TEMPLATE);
|
||||
if !imported {
|
||||
failed.insert(p.anchor);
|
||||
}
|
||||
} else {
|
||||
st.unsupported("a shared formula that is missing from the file");
|
||||
}
|
||||
}
|
||||
if shared_attaches >= MAX_SHARED_ATTACHES {
|
||||
st.unsupported("more shared formulas than the expansion cap; some formulas kept their values");
|
||||
}
|
||||
shared_attaches
|
||||
}
|
||||
|
||||
/// The other cells of a legacy array formula hold only cached results: dropped, as when an
|
||||
/// .xlsx is read, so that the formula fills its range on recalculation. If the formula itself
|
||||
/// couldn't be imported, they keep their values.
|
||||
fn blank_array_followers(st: &mut SheetState<'_>, pending: &[Pending]) {
|
||||
for p in pending {
|
||||
let filled = st.sheet.cell(p.anchor).and_then(|c| c.formula.as_ref()).and_then(|f| f.array).is_some_and(|r| r.contains(p.pos));
|
||||
if p.pos != p.anchor && filled {
|
||||
let style = st.sheet.style_id(p.pos);
|
||||
st.sheet.cells.set(p.pos, Cell { value: Value::Empty, formula: None, style });
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
fn worksheet(
|
||||
stream: &[u8],
|
||||
bs: &BoundSheet,
|
||||
@@ -591,8 +533,27 @@ fn worksheet(
|
||||
// Shared and array formulas: their definitions follow the first cell that uses them.
|
||||
let (arrays, shared) = (std::mem::take(&mut st.arrays), std::mem::take(&mut st.shared));
|
||||
let pending = std::mem::take(&mut st.pending);
|
||||
resolve_pending(&mut st, &arrays, &shared, &pending);
|
||||
blank_array_followers(&mut st, &pending);
|
||||
for p in &pending {
|
||||
if let Some((range, f)) = arrays.get(&p.anchor) {
|
||||
if p.pos == p.anchor {
|
||||
st.attach(p.pos, f, Mode::Cell, Some(*range));
|
||||
}
|
||||
} else if let Some(f) = shared.get(&p.anchor) {
|
||||
st.attach(p.pos, f, Mode::Shared, None);
|
||||
} else {
|
||||
st.unsupported("a shared formula that is missing from the file");
|
||||
}
|
||||
}
|
||||
// The other cells of a legacy array formula hold only cached results: dropped, as when an
|
||||
// .xlsx is read, so that the formula fills its range on recalculation. If the formula itself
|
||||
// couldn't be imported, they keep their values.
|
||||
for p in &pending {
|
||||
let filled = st.sheet.cell(p.anchor).and_then(|c| c.formula.as_ref()).and_then(|f| f.array).is_some_and(|r| r.contains(p.pos));
|
||||
if p.pos != p.anchor && filled {
|
||||
let style = st.sheet.style_id(p.pos);
|
||||
st.sheet.cells.set(p.pos, Cell { value: Value::Empty, formula: None, style });
|
||||
}
|
||||
}
|
||||
if st.dropped > 0 {
|
||||
report.warn(format!("{} cells outside the grid were skipped.", st.dropped));
|
||||
}
|
||||
|
||||
@@ -64,22 +64,17 @@ pub(super) struct Cursor<'a, 'b> {
|
||||
parts: &'b [&'a [u8]],
|
||||
part: usize,
|
||||
pos: usize,
|
||||
/// Bytes in every part after the current one, kept up to date as parts advance, so
|
||||
/// `remaining()` is constant-time however many `Continue` records a string session crosses.
|
||||
after: usize,
|
||||
}
|
||||
|
||||
impl<'a, 'b> Cursor<'a, 'b> {
|
||||
pub fn new(parts: &'b [&'a [u8]]) -> Self {
|
||||
let after = parts.iter().skip(1).map(|p| p.len()).sum::<usize>();
|
||||
Cursor { parts, part: 0, pos: 0, after }
|
||||
Cursor { parts, part: 0, pos: 0 }
|
||||
}
|
||||
|
||||
/// Skips to the next part when the current one is used up. Returns whether a new part began.
|
||||
fn settle(&mut self) -> bool {
|
||||
let mut moved = false;
|
||||
while self.parts.get(self.part).is_some_and(|p| self.pos >= p.len()) && self.part + 1 < self.parts.len() {
|
||||
self.after = self.after.saturating_sub(self.parts[self.part + 1].len());
|
||||
self.part += 1;
|
||||
self.pos = 0;
|
||||
moved = true;
|
||||
@@ -117,10 +112,10 @@ impl<'a, 'b> Cursor<'a, 'b> {
|
||||
Ok(())
|
||||
}
|
||||
|
||||
/// Bytes left in the record (all parts), constant-time in the part count.
|
||||
/// Bytes left in the record (all parts).
|
||||
pub fn remaining(&self) -> usize {
|
||||
let here = self.parts.get(self.part).map_or(0, |p| p.len().saturating_sub(self.pos));
|
||||
here.saturating_add(self.after)
|
||||
here + self.parts.iter().skip(self.part + 1).map(|p| p.len()).sum::<usize>()
|
||||
}
|
||||
|
||||
/// The rest of the record as one buffer.
|
||||
@@ -134,7 +129,6 @@ impl<'a, 'b> Cursor<'a, 'b> {
|
||||
}
|
||||
self.part = self.parts.len();
|
||||
self.pos = 0;
|
||||
self.after = 0;
|
||||
out
|
||||
}
|
||||
|
||||
@@ -201,23 +195,4 @@ mod tests {
|
||||
let mut c = Cursor::new(&parts);
|
||||
assert!(c.chars(MAX_CHARS + 1, false).is_err());
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn remaining_counts_every_part_wherever_the_cursor_is() {
|
||||
let parts: [&[u8]; 4] = [&[1, 2, 3], &[2, 3, 4, 5], &[3, 4, 5, 6, 7], &[4]];
|
||||
let mut c = Cursor::new(&parts);
|
||||
assert_eq!(c.remaining(), 13);
|
||||
for b in 1..=3u64 {
|
||||
assert_eq!(u64::from(c.u8().unwrap()), b, "part 0 in order");
|
||||
}
|
||||
assert_eq!(c.remaining(), 10, "after the first part is used up");
|
||||
for _ in 0..4 {
|
||||
c.u8().unwrap();
|
||||
}
|
||||
assert_eq!(c.remaining(), 6, "into the Continue parts");
|
||||
let mut c = Cursor::new(&parts);
|
||||
assert_eq!(u64::from(c.u8().unwrap()), 1);
|
||||
assert_eq!(c.rest().len(), 12);
|
||||
assert_eq!(c.remaining(), 0);
|
||||
}
|
||||
}
|
||||
|
||||
+4
-32
@@ -13,12 +13,6 @@ use crate::IoError;
|
||||
/// Maximum element nesting accepted in any part.
|
||||
pub const MAX_DEPTH: usize = 256;
|
||||
|
||||
/// Most elements kept in one part's parsed XML tree: an emptied `El` costs well over a hundred
|
||||
/// bytes, so a highly-compressible part of endless empty elements could fit the package byte
|
||||
/// limit and still take gigabytes of tree before any semantic check ran. Parts whose big runs
|
||||
/// matter (worksheet rows, shared strings) are streamed instead and never hold this tree.
|
||||
pub const MAX_XML_NODES: usize = 1_000_000;
|
||||
|
||||
/// An element: local name (namespace prefix dropped), attributes by local name, children and
|
||||
/// its direct text content.
|
||||
#[derive(Clone, Debug, Default, PartialEq)]
|
||||
@@ -229,27 +223,13 @@ pub fn parse_streaming(xml: &[u8], split: &[&str], cb: &mut dyn FnMut(El) -> Res
|
||||
let mut stack: Vec<El> = Vec::new();
|
||||
let mut root: Option<El> = None;
|
||||
let mut buf = Vec::new();
|
||||
// Elements kept in the tree so far, so a pathological part that passes the byte limits can
|
||||
// still be refused before the tree itself outgrows memory.
|
||||
let mut kept = 0usize;
|
||||
// Attach a finished element to its parent (or emit it / make it the root).
|
||||
fn finish(
|
||||
stack: &mut [El],
|
||||
root: &mut Option<El>,
|
||||
el: El,
|
||||
split: &[&str],
|
||||
cb: &mut dyn FnMut(El) -> Result<(), IoError>,
|
||||
kept: &mut usize,
|
||||
) -> Result<(), IoError> {
|
||||
fn finish(stack: &mut [El], root: &mut Option<El>, el: El, split: &[&str], cb: &mut dyn FnMut(El) -> Result<(), IoError>) -> Result<(), IoError> {
|
||||
match stack.last_mut() {
|
||||
Some(parent) => {
|
||||
if split.contains(&el.name.as_str()) {
|
||||
cb(el)?;
|
||||
} else {
|
||||
*kept = kept.saturating_add(1);
|
||||
if *kept > MAX_XML_NODES {
|
||||
return Err(IoError::Xml(format!("the part holds more than {MAX_XML_NODES} elements")));
|
||||
}
|
||||
parent.children.push(el);
|
||||
}
|
||||
}
|
||||
@@ -272,11 +252,11 @@ pub fn parse_streaming(xml: &[u8], split: &[&str], cb: &mut dyn FnMut(El) -> Res
|
||||
}
|
||||
Event::Empty(e) => {
|
||||
let el = start_el(&e);
|
||||
finish(&mut stack, &mut root, el, split, cb, &mut kept)?;
|
||||
finish(&mut stack, &mut root, el, split, cb)?;
|
||||
}
|
||||
Event::End(_) => {
|
||||
if let Some(el) = stack.pop() {
|
||||
finish(&mut stack, &mut root, el, split, cb, &mut kept)?;
|
||||
finish(&mut stack, &mut root, el, split, cb)?;
|
||||
}
|
||||
}
|
||||
Event::Text(t) => {
|
||||
@@ -315,7 +295,7 @@ pub fn parse_streaming(xml: &[u8], split: &[&str], cb: &mut dyn FnMut(El) -> Res
|
||||
}
|
||||
// Unclosed elements (truncated part): close them so we keep what we have.
|
||||
while let Some(el) = stack.pop() {
|
||||
finish(&mut stack, &mut root, el, split, cb, &mut kept)?;
|
||||
finish(&mut stack, &mut root, el, split, cb)?;
|
||||
}
|
||||
root.ok_or_else(|| IoError::Xml("no root element".into()))
|
||||
}
|
||||
@@ -495,14 +475,6 @@ mod tests {
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_part_of_endless_empty_elements_is_refused() {
|
||||
assert!(parse(b"<r><x/><y t=\"1\"/><x/></r>").is_ok());
|
||||
let xml = format!("<r>{}<y/></r>", "<x/>".repeat(MAX_XML_NODES + 1));
|
||||
let err = parse(xml.as_bytes()).unwrap_err();
|
||||
assert!(err.to_string().contains("elements"), "{err}");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn phonetic_data_validates_offsets() {
|
||||
let x = r#"<si><t>課きく</t><rPh sb="0" eb="1"><t>カ</t></rPh><rPh sb="2" eb="9"><t>ク</t></rPh><rPh sb="2" eb="2"><t>x</t></rPh><rPh sb="1"/></si>"#;
|
||||
|
||||
@@ -17,16 +17,12 @@ if [[ -z "${APPLE_CERTIFICATE:-}" ]]; then
|
||||
exit 0
|
||||
fi
|
||||
|
||||
BASE="${RUNNER_TEMP:-${TMPDIR:-/tmp}}"
|
||||
# A private directory created exclusively, not predictable names in a possibly shared temp
|
||||
# root: another local user's symlink at `gridcraft-signing.p12` would otherwise select the
|
||||
# write target and capture the decoded certificate.
|
||||
WORK="$(mktemp -d "$BASE/gridcraft-signing.XXXXXXXX")"
|
||||
KEYCHAIN="$WORK/gridcraft-signing.keychain-db"
|
||||
CERT="$WORK/gridcraft-signing.p12"
|
||||
TMP="${RUNNER_TEMP:-${TMPDIR:-/tmp}}"
|
||||
KEYCHAIN="$TMP/gridcraft-signing.keychain-db"
|
||||
CERT="$TMP/gridcraft-signing.p12"
|
||||
KC_PASS="${KEYCHAIN_PASSWORD:-$(openssl rand -hex 24)}"
|
||||
|
||||
trap 'rm -rf "$WORK"' EXIT
|
||||
trap 'rm -f "$CERT"' EXIT
|
||||
printf '%s' "$APPLE_CERTIFICATE" | base64 --decode >"$CERT"
|
||||
security create-keychain -p "$KC_PASS" "$KEYCHAIN"
|
||||
security set-keychain-settings -lut 21600 "$KEYCHAIN"
|
||||
|
||||
+3
-24
@@ -2,9 +2,6 @@
|
||||
|
||||
use std::path::Path;
|
||||
|
||||
#[cfg(unix)]
|
||||
use std::os::unix::fs::MetadataExt as _;
|
||||
|
||||
#[derive(Default, Clone, Copy)]
|
||||
struct Counts {
|
||||
files: usize,
|
||||
@@ -13,24 +10,7 @@ struct Counts {
|
||||
proptests: usize,
|
||||
}
|
||||
|
||||
fn walk(dir: &Path, c: &mut Counts, seen: &mut std::collections::HashSet<u64>, budget: &mut usize) {
|
||||
if *budget == 0 {
|
||||
// Hard stop for pathological trees: two directory symlinks pointing at each other
|
||||
// would otherwise be walked exponentially.
|
||||
return;
|
||||
}
|
||||
// Directory identities (dev+inode on unix) make a symlink to a visited directory a
|
||||
// no-op, not another exponential branch of the scan.
|
||||
#[cfg(unix)]
|
||||
let id = std::fs::metadata(dir).ok().and_then(|m| m.dev().checked_mul(1 << 60).and_then(|d| d.checked_add(m.ino())));
|
||||
#[cfg(not(unix))]
|
||||
let id: Option<u64> = None;
|
||||
if let Some(id) = id
|
||||
&& !seen.insert(id)
|
||||
{
|
||||
return;
|
||||
}
|
||||
*budget -= 1;
|
||||
fn walk(dir: &Path, c: &mut Counts) {
|
||||
let Ok(rd) = std::fs::read_dir(dir) else {
|
||||
return;
|
||||
};
|
||||
@@ -38,11 +18,10 @@ fn walk(dir: &Path, c: &mut Counts, seen: &mut std::collections::HashSet<u64>, b
|
||||
let p = e.path();
|
||||
let name = e.file_name();
|
||||
let name = name.to_string_lossy();
|
||||
// A symlinked directory is descended only if its identity is new (checked above).
|
||||
if p.is_dir() {
|
||||
// Nested fuzz workspaces and build output are not part of the crate.
|
||||
if name != "target" && name != "fuzz" && !name.starts_with('.') {
|
||||
walk(&p, c, seen, budget);
|
||||
walk(&p, c);
|
||||
}
|
||||
} else if name.ends_with(".rs")
|
||||
&& let Ok(s) = std::fs::read_to_string(&p)
|
||||
@@ -92,7 +71,7 @@ pub fn run(_root: &Path, exact: bool) -> Result<(), String> {
|
||||
continue;
|
||||
};
|
||||
let mut c = Counts::default();
|
||||
walk(&dir, &mut c, &mut std::collections::HashSet::new(), &mut 10_000);
|
||||
walk(&dir, &mut c);
|
||||
let ex = if exact { exact_tests(&name) } else { None };
|
||||
rows.push((name, c, ex));
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user