[NopeCharon]

What and why

Git leaves paths that only contain spaces unquoted, so the header for my notes.txt is diff --git a/my notes.txt b/my notes.txt. _paths_from_diff_header split it with shlex.split and took the 3rd and 4th words, so the paths became a/my and notes.txt.

Ordinary text edits hid this because the following --- a/... / +++ b/... lines overwrite the path. Deleted files (+++ /dev/null), binary files and mode-only changes have no such line. The canvas named them notes.txt and then appended the real path from changed_files as a second, hunk-less "modified" entry (details in #280).

Without a rename or copy, both names in the header are the same path behind their prefixes. That is also how Git itself splits an unquoted header. For unquoted headers, the parser now splits at the space where the two names match after removing their first path component, or where they are identical (diff.noprefix). Renames and copies keep the previous fallback and are still corrected by their rename from / rename to lines. Quoted headers are unchanged.

Fixes #280

Linked Issue for product-semantic or public-contract changes: none (routine fix)

Change class

• Routine fix, documentation, test, maintenance, or presentation-only UI
• Product-semantic or public-contract change discussed in the linked Issue
• Isolated, default-off experiment

Owning layer: server/unified_diff.py, the parser that turns git diff output into the provider-neutral diff canvas shape.

User-visible effect, or none: in the Work attempt diff canvas, deleted, binary and mode-only files whose paths contain spaces keep their real path and are listed once.

Compatibility or migration impact, or none: none. The output shape is unchanged.

Evidence

Commands and manual journeys run:

• python -m pytest -q tests/test_unified_diff_canvas.py tests/test_work_export_integration.py: 29 passed (Linux, uv sync --locked --extra dev). python tests/test_unified_diff_canvas.py also passes.

• The new test test_parse_unified_diff_keeps_spaced_paths_without_marker_lines fails on main and passes with this change.

• python -m ruff check server tests/test_unified_diff_canvas.py: passed.

• Parsed real git diff HEAD output from a scratch repository after deleting my file.txt and editing the binary my img.bin. Before: file.txt (deleted), img.bin, plus duplicate my file.txt / my img.bin entries. After: my file.txt (deleted) and my img.bin only.

• Relevant Python tests pass

• CPU/model-less baseline remains supported

• Electron npm run build passes when Electron code changed

• Dependency audit passes when dependencies changed

• Before/after screenshots are attached for visible UI changes

• Documentation/examples are updated for changed settings or contracts

Final check

• This PR addresses one coherent problem without unrelated cleanup
• It does not add a speculative API, fallback, or compatibility path
• No secrets, local state, model weights, voice material, or restricted assets are included
• Third-party notices and provenance are preserved

[Lucas1479]

Thank you for the precise path-parsing fix and regression test. Independent real-Git checks confirm that deleted, binary and mode-only files with spaces retain their names and no longer produce duplicate canvas entries. The branch has been updated against main; the original patch is unchanged, and merge will wait for CI on this updated head. The comment about copy headers is broader than the current parser supports, but that is a separate existing copy-detection case and does not block this fix.