Repository navigation
fix(desktop): resolve Save sheet controls to their parent window - #382
Merged
Merged
Conversation
Pin desktop-tools ccb4750, which adds peekaboo-sheet-window.patch. Sheet controls link to the sheet (or, for Open/Save panels, the Open and Save Panel Service's window), so the identity read before focus and value writes found no observed window and refused every mutation. Their owner is now the nearest same-process AXWindow. Fixes #378 Why: package.json | desktop-tools must stay at or after ccb4750 so sheet controls resolve to the AXWindow the sheet is attached to; older pins refuse every mutation on Save/Open sheets (#378)
This was referenced Oct 15, 2026
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #378. Pins
@githubnext/desktop-toolsto ccb4750 (githubnext/desktop-tools#8). That commit addspeekaboo-sheet-window.patch.Cause. Before a focus or value write, Peekaboo's
ActionInputDriver.observationTargetreads the element's identity throughDetachedExactWindowFocusReader. That read took its CGWindowID only from the element'sAXWindowlink, and any failure there produced "The native Accessibility target could not be revalidated before mutation." A Save sheet is its own WindowServer window but is missing from TextEdit's AXWindows, so the observation reaches it only through the document window's tree. The sheet's controls link to the sheet. An Open or Save panel's controls resolve to the Open and Save Panel Service's own window: trycua/cua#4507 recorded the chain button → service window 173, AXSheet 172, parent AXWindow 167. Neither can be resolved to the observed document window, so every click and type on the sheet was refused before dispatch.Change (in desktop-tools).
AXWindowIDResolver.owningWindowIDaccepts the AXWindow link only when it is a same-processAXWindow. Otherwise it follows a boundedAXParentchain to the nearest same-processAXWindow. An element with no such ancestor keeps its direct window identity. The detached reader uses the same helper within its deadline, so element identity, exact-window click receipts and post-write observation all name the same window. Sibling windows are still refused.Verification. Done:
bun types,bun run ci, the native desktop-tools build run the wayapps/desktop/scripts/helper.tsruns it (the old stack prefix plus the new patch applied and built), desktop-tools CI (unsigned macOS build), and the updated Peekaboo unit testsAXWindowIDResolverTestsandDetachedExactWindowFocusMetadataTests, which use injected fakes and no live AX.Unverified against a live Save sheet. This Mac was running a study, so no GUI app could be opened. The study lead will rerun scenario 20 step 3. If it still refuses, the next suspect is the reader's 50 ms per-call AX messaging cap on the panel's remote elements.
Built in Ace
No — built with Claude Code outside Ace by an overnight issue watcher, because this Mac's Ace Canary is a participant in the running two-person study (#218); dogfooding gap tracked in #5.