`executeCommand("undo")` is non-functional in some headless `.vscode-test`
instances (it does not restore the buffer — see #54's diagnosis), which
false-fails every undo-dependent host E2E and turns `main`'s E2E red there even
though the product code is fine.
Add a RUNTIME preflight probe (`undoCapable.ts`): edit a scratch buffer, undo,
and report whether the buffer was actually restored (memoized per run). The #38
undoMarks suite gates on it via `suiteSetup` — running normally where undo works
(full coverage preserved, real regressions still caught) and skipping with a LOUD
console warning where it doesn't (no false red, no silent loss — the skip is
logged and shows as `pending`).
This fixes the "main E2E red" symptom. The underlying headless-undo limitation is
documented in #54; #40's undo-behavior coverage runs wherever undo works.
Test-infra only — no product code changed. Verified: with undo broken locally the
#38 suite skips (1 pending) and both E2E passes exit 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Undo in the editor rendered wrong marks in the F10 review preview: restored
baseline text (and reverted Claude text) was colored as human-authored. Root
cause — attributionController.onDidChange attributed every non-seam document
change to currentAuthor() (always human) and ignored e.reason, so an undo/redo
that re-inserts text created a fresh human span over it. renderReview is pure;
the wrong marks came from these false author spans.
Fix: on e.reason === Undo|Redo, reconcile span geometry but do NOT attribute the
re-inserted chars — an undo is history navigation, not authorship, so restored
text stays neutral (unattributed) rather than falsely claimed by the human.
applyChange gains an `attributeInserted` flag (default true; false on undo/redo);
seam matching is also skipped on undo/redo (the seam only applies forward edits).
Restored text becomes neutral rather than recovering its exact prior provenance
(e.g. undoing a deletion of Claude's text shows it unattributed, not blue) —
perfect restoration would need an attribution history stack synced to the editor
undo stack (fragile due to edit coalescing); filed mentally as a follow-up. The
neutral behavior removes the misleading marks, which is the reported defect.
Tests: E2E reproduces the mid-edit-undo case (buffer stays dirty so the disk-sync
guard doesn't mask it) and asserts restored text is unattributed; +3 unit tests
for the geometric-only applyChange path. 197 unit + 50 E2E green; typecheck clean.
Closes#38
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ben Stull