address review: run MP4 table guard only for MP4 or unknown containers

crates/engine/src/file_access.rs

@@ -85,20 +85,14 @@ impl AutomationRoots {
 
     /// May the file or folder at `path` be read (or written)? `Err` says why not.
     pub fn check(&self, path: &str, access: Access) -> Result<(), String> {
-        self.check_path(Path::new(path), access)
-    }
-
-    /// [`Self::check`] for a path that may not be UTF-8: the checked path is the one opened.
-    pub fn check_path(&self, path: &Path, access: Access) -> Result<(), String> {
-        let path_text = path.display();
         let roots = match access {
             Access::Read => &self.read,
             Access::Write => &self.write,
         };
         if roots.is_empty() {
             return match access {
-                Access::Read => Err(format!("automation filesystem access is not granted: read authority is absent: {path_text}")),
-                Access::Write => Err(format!("automation filesystem access is not granted: write authority is absent: {path_text}")),
+                Access::Read => Err(format!("automation filesystem access is not granted: read authority is absent: {path}")),
+                Access::Write => Err(format!("automation filesystem access is not granted: write authority is absent: {path}")),
             };
         }
         let resolved = resolve(path)?;
@@ -107,8 +101,8 @@ impl AutomationRoots {
         }
         let roots = roots.iter().map(|root| root.shown.display().to_string()).collect::<Vec<_>>().join(", ");
         match access {
-            Access::Read => Err(format!("automation path rejected: outside the read roots {roots}: {path_text}")),
-            Access::Write => Err(format!("automation path rejected: outside the write roots {roots}: {path_text}")),
+            Access::Read => Err(format!("automation path rejected: outside the read roots {roots}: {path}")),
+            Access::Write => Err(format!("automation path rejected: outside the write roots {roots}: {path}")),
         }
     }
 }
@@ -197,23 +191,17 @@ pub fn check_write(path: &str) -> Result<(), String> {
     current().map_or(Ok(()), |r| r.check(path, Access::Write))
 }
 
-/// May `path` be written here? Always, without roots in force. For a path that may not be UTF-8.
-pub fn check_write_path(path: &Path) -> Result<(), String> {
-    current().map_or(Ok(()), |r| r.check_path(path, Access::Write))
-}
-
 /// `path` made absolute with its links followed; for a path that doesn't exist (yet), its nearest
 /// existing folder's links followed and the rest appended.
-fn resolve(path: &Path) -> Result<PathBuf, String> {
-    if path.as_os_str().is_empty() {
+fn resolve(path: &str) -> Result<PathBuf, String> {
+    if path.is_empty() {
         return Err("automation path rejected: path is empty".into());
     }
-    let shown = path.display();
     // Checked as given: making `…\NUL` absolute gives the device path `\\.\NUL` on Windows.
-    if cfg!(windows) && path.components().any(|c| matches!(c, Component::Normal(n) if is_device_name(&n.to_string_lossy()))) {
-        return Err(format!("automation path rejected: reserved device names are not allowed: {shown}"));
+    if cfg!(windows) && Path::new(path).components().any(|c| matches!(c, Component::Normal(n) if is_device_name(&n.to_string_lossy()))) {
+        return Err(format!("automation path rejected: reserved device names are not allowed: {path}"));
     }
-    let absolute = std::path::absolute(path).map_err(|e| format!("automation path rejected: cannot resolve {shown}: {e}"))?;
+    let absolute = std::path::absolute(path).map_err(|e| format!("automation path rejected: cannot resolve {path}: {e}"))?;
     // The names below the nearest existing folder, last first.
     let mut missing: Vec<OsString> = vec![];
     let mut at = absolute.as_path();
@@ -224,11 +212,11 @@ fn resolve(path: &Path) -> Result<PathBuf, String> {
                 return Ok(found);
             }
             Err(e) => match std::fs::symlink_metadata(at) {
-                Ok(m) if m.file_type().is_symlink() => return Err(format!("automation path rejected: a link that leads nowhere: {shown}")),
+                Ok(m) if m.file_type().is_symlink() => return Err(format!("automation path rejected: a link that leads nowhere: {path}")),
                 // There, yet its real path can't be had (no permission): nothing to compare.
-                Ok(_) => return Err(format!("automation path rejected: cannot resolve {shown}: {e}")),
+                Ok(_) => return Err(format!("automation path rejected: cannot resolve {path}: {e}")),
                 Err(e) if e.kind() == std::io::ErrorKind::NotFound => {}
-                Err(e) => return Err(format!("automation path rejected: cannot resolve {shown}: {e}")),
+                Err(e) => return Err(format!("automation path rejected: cannot resolve {path}: {e}")),
             },
         }
         let parent = at.parent();
@@ -238,12 +226,12 @@ fn resolve(path: &Path) -> Result<PathBuf, String> {
                 at = parent;
             }
             (Some(Component::ParentDir), _) => {
-                return Err(format!("automation path rejected: parent traversal below a folder that doesn't exist: {shown}"));
+                return Err(format!("automation path rejected: parent traversal below a folder that doesn't exist: {path}"));
             }
             // Down to the root (or drive) and it doesn't exist either.
             _ => {
                 let e = std::io::Error::from(std::io::ErrorKind::NotFound);
-                return Err(format!("automation path rejected: cannot resolve {shown}: {e}"));
+                return Err(format!("automation path rejected: cannot resolve {path}: {e}"));
             }
         }
     }

crates/engine/src/spell.rs

@@ -668,8 +668,6 @@ pub fn add_user_word(word: &str) -> std::io::Result<()> {
         return Ok(());
     }
     let path = user_dictionary_path().ok_or_else(|| std::io::Error::new(std::io::ErrorKind::NotFound, "no folder for the user dictionary"))?;
-    // A confined automation client may write only inside its write roots, this file included.
-    crate::file_access::check_write_path(&path).map_err(|e| std::io::Error::new(std::io::ErrorKind::PermissionDenied, e))?;
     if let Some(d) = path.parent() {
         std::fs::create_dir_all(d)?;
     }
@@ -955,20 +953,6 @@ SFX D Y 1\nSFX D 0 ed/S .\n";
         assert_eq!(w, vec!["check", "done"]);
     }
 
-    #[test]
-    fn a_confined_client_cannot_write_the_user_dictionary_outside_its_roots() {
-        let root = std::env::temp_dir().join(format!("deckcraft-dict-root-{}", std::process::id()));
-        std::fs::create_dir_all(&root).unwrap();
-        let roots = crate::file_access::AutomationRoots::new(std::slice::from_ref(&root), std::slice::from_ref(&root)).unwrap();
-        // The user dictionary is a file of its own outside `root`, not whatever HOME points at.
-        let outside = std::env::temp_dir().join(format!("deckcraft-dict-outside-{}", std::process::id()));
-        set_user_dictionary_path(outside.join("user.dic"));
-        let r = crate::file_access::confine(roots.as_ref(), || add_user_word("zqxconfinedword"));
-        assert_eq!(r.map_err(|e| e.kind()), Err(std::io::ErrorKind::PermissionDenied));
-        assert!(!outside.exists(), "the refused write created nothing");
-        let _ = std::fs::remove_dir_all(&root);
-    }
-
     #[test]
     fn long_and_numeric_flags_and_latin1() {
         let d = Dictionary::parse(b"FLAG long\nSFX Aa Y 1\nSFX Aa 0 s .\n", b"1\ncat/Aa\n");

crates/media/src/audio.rs

@@ -81,7 +81,7 @@ pub fn decode(bytes: &Bytes) -> Result<Pcm> {
             Some(r) => r,
             None => contained(|| decode_symphonia(bytes, "mkv")),
         },
-        Container::Mp4 if !crate::mp4guard::tables_fit(bytes) => Err(MediaError::Corrupt("the MP4's sample tables do not fit the file".into())),
+        _ if !crate::mp4guard::probe_fits(bytes) => Err(MediaError::Corrupt("the MP4's sample tables do not fit the file".into())),
         c => contained(|| decode_symphonia(bytes, c.name())),
     }
 }

crates/media/src/mp4guard.rs

@@ -12,6 +12,20 @@ pub(crate) fn tables_fit(b: &[u8]) -> bool {
     boxes_fit(b, b.len() as u64, 0)
 }
 
+/// Bytes symphonia's probe scans for a format marker before giving up.
+const PROBE_SEARCH_LIMIT: usize = 1024 * 1024;
+
+/// Whether symphonia can be given `b`, whatever the sniffer made of it: its probe takes the first
+/// `ftyp` within its search window as the start of an MP4 (the size field 4 bytes before it), so
+/// the table check must start where that parser will, not at offset 0.
+pub(crate) fn probe_fits(b: &[u8]) -> bool {
+    let head = &b[..b.len().min(PROBE_SEARCH_LIMIT + 16)];
+    match head.windows(4).position(|w| w == b"ftyp") {
+        Some(p) => tables_fit(&b[p.saturating_sub(4)..]),
+        None => true,
+    }
+}
+
 fn be32(b: &[u8], at: usize) -> Option<u64> {
     b.get(at..at.checked_add(4)?).map(|s| u32::from_be_bytes([s[0], s[1], s[2], s[3]]) as u64)
 }
@@ -74,3 +88,29 @@ fn table_fits(kind: &[u8], body: &[u8], file_len: u64) -> bool {
     };
     be32(body, 4).is_some_and(|count| count <= body.len().saturating_sub(8) as u64 / entry)
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// An MP4 whose `stsc` count asks for far more entries than the file holds.
+    fn mp4_with_huge_stsc() -> Vec<u8> {
+        let stsc = [&24u32.to_be_bytes()[..], b"stsc", &[0, 0, 0, 0], &u32::MAX.to_be_bytes(), &[0; 8]].concat();
+        let stbl = [&(stsc.len() as u32 + 8).to_be_bytes()[..], b"stbl", &stsc].concat();
+        let moov = [&(stbl.len() as u32 + 8).to_be_bytes()[..], b"moov", &stbl].concat();
+        [&16u32.to_be_bytes()[..], b"ftyp", b"isom", &[0; 4], &moov].concat()
+    }
+
+    #[test]
+    fn a_prefixed_mp4_is_checked_where_symphonia_starts_it() {
+        let mp4 = mp4_with_huge_stsc();
+        assert!(!tables_fit(&mp4));
+        assert!(!probe_fits(&mp4));
+        let prefixed = [&b"junk before the file "[..], &mp4].concat();
+        assert_eq!(crate::Container::sniff(&prefixed), crate::Container::Unknown);
+        assert!(tables_fit(&prefixed), "offset 0 sees no MP4 boxes");
+        assert!(!probe_fits(&prefixed));
+        let far = [&vec![0u8; PROBE_SEARCH_LIMIT + 16][..], &mp4].concat();
+        assert!(probe_fits(&far), "past the probe window symphonia never reads it as an MP4");
+    }
+}

crates/media/src/probe.rs

@@ -71,7 +71,8 @@ pub fn probe_headers(bytes: &Bytes) -> Result<MediaInfo> {
     };
     if info.audio.is_none() && info.video.is_none() || info.audio.as_ref().is_some_and(|a| a.sample_rate == 0) {
         // Plain audio files (and anything our demuxers didn't describe): symphonia's probe.
-        if c == Container::Mp4 && !crate::mp4guard::tables_fit(bytes) {
+        // Known non-MP4 containers (a WAV's audio bytes can spell an `ftyp`) skip the MP4 table check.
+        if matches!(c, Container::Mp4 | Container::Unknown) && !crate::mp4guard::probe_fits(bytes) {
             return Err(MediaError::Corrupt("the MP4's sample tables do not fit the file".into()));
         }
         let (codec, rate, channels, frames) = std::panic::catch_unwind(|| crate::audio::probe_symphonia(bytes, c.name()))