Repository navigation
fix: backport five user-reported fixes to 0.12.1 (batch 30) - #32194
Conversation
* fix(chat): label Windows attachment paths by file name Sent-message attachment cards, inline file chips and the pinned prompt preview on a Windows gateway showed the whole native path, because the labels split paths on `/` alone and a path like `C:\Users\me\report.docx` was one segment. buildFileLabels splits a Windows-shaped path (drive-rooted or UNC) on both `\` and `/`, using the same backslashIsSeparator rule as pathBasename, so a POSIX path keeps `\` inside its file name. Same- named files still widen by parent folder until unique, and labels join with `/` like the composer's. A chip whose path is missing from the attachment list falls back to its file name; a label another chip already holds uses the full path instead, and optimistic-mention recovery leaves a marker's own entry in place. The embedded transcript tool-pill button and promptPreview use pathBasename for the same reason. Fixes #27796 * fix(chat): show a chipped attachment once in a sent message A mixed-separator Windows path such as `C:/repo\src\a.ts` labels its inline chip `@a.ts`, but the unreferenced-attachment check reads the rewritten text and splits paths on `/` only, so it missed that chip and drew a second card for the same file. A file the message already shows as an inline chip now counts as referenced, so it gets no extra card. Part of #27796 --------- Co-authored-by: Matt Briones <mdbrio@amazon.com> (cherry picked from commit 2c109f6)
…1215) An app's extension-scoped file-menu row matched a Windows dotfile (C:\repo\.gitignore for "gitignore") because fileMenuItemMatches took the name with a /-only split, so the whole native path came through and its last dot sat past index 0. The app-SDK tool pill's open-file chip showed the whole C:\... path for the same reason. Both sites now take the name through pathBasename, which is byte-identical to the old split for every POSIX path, and both files join the windowsBasenameSites source scan. Closes #27792 (cherry picked from commit 1e19151)
…31205) On Windows the AWS CLI reports a credential_process program it cannot start as "[WinError 2] The system cannot find the file specified", with no program name. The detector only read the POSIX "[Errno 2] ... '<name>'" text, so cloud doctor told the user to re-run SSO setup instead of giving the helper an absolute path. When the stderr carries that Windows text, read the program from the profile's own credential_process setting (aws configure get): a leading double-quoted program is read whole, spaces included, as the CLI on Windows reads it, otherwise its first word. It keeps the bare-name rule the POSIX detector keeps: a name with a path separator, an unset value, or an agent session that may not run the call keeps today's generic note. (cherry picked from commit 0eecd84)
…#31375) The chat watchdog's model-wait branch deferred every WORKING reading with no upper bound. That reading is CPU or IO movement anywhere in the runtime's process tree, which a busy MCP server also produces, so a turn that streamed nothing was never probed until the prompt timeout. Defer it only while the stale idle is within watchdog.tool_stall_hard_cap_secs, as the tool branch already does for its tree-wide readings, then send the existing non-lethal stale probe, labelled window=working_cap. Closes #24655 (cherry picked from commit db1c79b) Release backport: docs/system-specs/modules/metrics.md conflicted in one table row; resolved by keeping release's text and applying this change's rewrite of the working_cap phrase. Code, config and tests applied unchanged. Original author: Bolin Chen (#31375).
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: this PR copies five fixes already merged on Notes
[DESIGN-REVIEWED] 6a38cb4 Description read[DESCRIPTION-READ] 2fea3b343e1cefe4484420fffd42f78b9939d46c40932ba4bb3db6ea454dc3cf That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
UX Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)UX-level review of UX-Verdict: PASS Cannot evaluate: missing screenshots of the Windows attachment and file-chip labels this backport changes. No image is attached, so the blind read was not performed. Evidence gaps
[UX-REVIEWED] 6a38cb4 Description read[DESCRIPTION-READ] 6893bed6ef226a2173c49da0cbc34c7a5a89f467d6534ca252268e4dd745f94f That sha256 names everything this verdict read: the PR description, plus the 3 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks: this is five cherry-picks of fixes already merged on What this change shipsInventory (8 items) — 8 justifiedIntent: ship five user-reported fixes already merged on
[FIRST-PRINCIPLES-REVIEWED] 6a38cb4 Description read[DESCRIPTION-READ] 2683769e3de89bca11d1d6b2a07ee3ed88dae30eeb68932818374ef07de91a69 That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
Problem / Motivation
Goal: Bring five small user-reported fixes that merged on
mainafterv0.12.0ontorelease/0.12.0, one commit per fix, so the 0.12.1 patch ships them and each one can be reverted alone.Why it matters
/only.kirocrew cloud doctornever names a missingcredential_processhelper.Not a goal
main; nothing is changed only here.CHANGELOG.mdchange. Cuttingv0.12.1-rc.Nand the bare tag follow docs/build/release.md and stay with the maintainers.What changed (motivation → approach → change)
Each main squash commit was cherry-picked with
-xontorelease/0.12.0@4d9f6b123e, in merge order, one commit each.2c109f6117e776f64f18041f98d623f807a2549a1e191512edd8f66bcc560eecd845d10fc18a14b7db1c79baca6a38cb495dThe one hand-resolved hunk is documentation. In #31375,
docs/system-specs/modules/metrics.mdconflicted in one table row; it was resolved by keeping this branch's row and applying #31375's rewrite of itsworking_capphrase, and the result is identical tomain's line. #31375's other changes (watchdog code, two config help texts andconfig-baseline.json) applied with no conflict; its config edits change help text only, not a setting. That commit is authored with the public identity the push guard requires and names Bolin Chen as the original author in its message; the other four keep their original authors. Every pick's added and removed lines are identical to its main commit.#28985 comes before #31215 because #31215's tests expect
app-sdk/messageRenderers.tsxto usepathBasename, which #28985 adds. #31205 reads the AWScredential_processsetting only to name a missing helper incloud doctor; it changes no credential handling.builder-security checked: no pick changes a bind, a listener, a route, an auth check, an ingress rule or a storage grant.
Backwards compatibility
Compatible: each change is the one already merged on
main, and each PR's own compatibility note applies unchanged.Removes nothing: no pick hides, deletes or renames anything 0.12.0 accepts.
Tests
Proof: not provable -- other: a backport of main commits whose proofs ran on their own PRs
release/*, so the related set was run here: every touched test file plus every test file that imports a touched module or reads the touched config, 47 files, one pytest process per batch of 20,-n 0. 2631 passed. 1 failed:test_file_line_cap.py::TestTheCheckout::test_the_change_keeps_the_cap_and_its_baseline_rows, which fails the same way on a cleanrelease/0.12.0checkout becausedashboard/handlers/agents.pyis 2,002 lines there; refactor: backport the agents.py effort-levels split to 0.12.1 (batch 29) #32184 clears it.windowsBasenameSites.test.tsx,AppSdkMessageRenderersCov80.test.tsx, and every test that reads a touched module, 46 files, 1000 passed.Manual verification
N/A - every pick was verified on its own main PR. The 0.12.1 RC soak is the end-to-end check.
Screenshots / video
Why no screenshot: each frontend change here is applied from
mainunchanged; its look is reviewed on its own PR, not here.Related Issues
Backport to the 0.12 line for 0.12.1 of #28985, #30999, #31215, #31205 and #31375. Source issues: #27796, #16574, #27792, #27866, #24655.
Sibling sweep
Siblings: none -- each source PR carried its own sibling sweep; this PR only carries those fixes.
Pattern harvest
Not generalizable: this PR only carries fixes already on
main. Each source PR has its own harvest.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.