Task 4 gated every surface on CoeditingRegistry but only wired
ThreadController/EditorProposalController to registry.onDidChange for
restore-on-enter — ProposalController.renderAll and
AttributionController.loadAll were never called on the enter transition, so
a document's FIRST enter with pre-existing sidecar proposals/attribution
showed threads but not proposal decorations/CodeLens or committed
author-coloring until a later edit happened to fire them (a gap Task 4's
own report flagged as known and unfixed).
Both controllers now subscribe to registry.onDidChange in their
constructors and call renderAll/loadAll on entry only (both already
self-gate on isCoediting; exit stays hide-only, mirroring
ThreadController). Construction order in extension.ts (attribution before
proposals before EditorProposalController) keeps the same-tick read order
correct.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Task 2 of the native-surfaces migration plan. GitBaselineAdapter resolves a
document's HEAD blob via the built-in vscode.git extension (hardened with a
.git/logs/HEAD watch that nudges repo.status(), since the git extension's own
watcher doesn't reliably notice out-of-band commits on non-workspace-folder
repos in the E2E host). DiffViewController is reworked into a baseline router:
head mode (git-tracked, never persisted, re-read on commit) vs snapshot mode
(captured on CoeditingRegistry entry, re-pinned by "Mark Changes as Reviewed"
— renamed from cowriting.pinDiffBaseline, D14). The shipped machine-landing
baseline advance (#48/INV-18) is retired: a landed Claude edit now stays a
visible change-since-baseline until commit or review (INV-7/D21). Legacy
on-disk reasons ("opened"/"machine-landing") migrate to "entered" via
BaselineStore.normalizeReason on load.
Co-Authored-By: Claude Fable 5 <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>
Code-quality follow-up: with editor decorations gone (Task 1), the visibility
toggle was dead/misleading — nothing reads `visible`. Remove the field,
toggle()/isVisible(), the cowriting.toggleAttribution registration, and its
package.json command + palette declarations (spec §6.2 retires it). renderActive/
render stay — they refresh the orphan status bar on editor switch.
E2E toggle assertions (attribution.test.ts, noWorkspace.test.ts) are updated in SLICE-4.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
cowriting.editSelection collapsed four distinct failures (no editor / no
selection / unsaved / outside the workspace folder) into one misleading
"select some text in a workspace document first" warning — so editing a
selection in a file OUTSIDE the EDH workspace root (e.g. a content-repo file
while the EDH opens sandbox/) was rejected as if no text were selected.
- New pure, vscode-free src/workspacePath.ts: isUnderRoot() (separator-bounded
membership, fixing a latent startsWith() prefix collision where a sibling
whose name prefixes the root — e.g. vscode-cowriting-plugin-content vs
vscode-cowriting-plugin — falsely matched) + selectionRejection() (one
message per condition). Unit-tested (8 cases), incl. the reported bug.
- Wire isUnderRoot into all five membership checks: extension.ts (editSelection
guard + renderIfOpen) and the thread/attribution/proposal controllers.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
F6 §6.2/§6.4, INV-9/INV-18. Additive EventEmitter on the seam owner, fired
after a real (non-no-op) applyEdit succeeds. Seam signature/return unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The accept E2E exposed a latent F3 limitation: VS Code word-diffs one
applied WorkspaceEdit into several minimal hunks when old/new share
interior tokens, so the registry's per-hunk exact match missed and seam
edits fell through as fragmented human spans. PendingEditRegistry.match
is replaced by matchEvent (all hunks inside the registered full range +
equal net delta — one applyEdit is one change event). Plan AMENDMENT 1.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
BenStullsBets
benstull
Ben Stull