Skip to content

fix(cm): integrate reviewed V2 reliability fixes - #96

Merged
trmosala merged 54 commits into
CM-Split-v2from
v2-verified-fixes
Oct 21, 2026
Merged

trmosala merged 54 commits into
CM-Split-v2from
v2-verified-fixes

Conversation

@trmosala

@trmosala trmosala commented Oct 20, 2026 •

Copy link
Copy Markdown
Owner

Issue for this PR

Addresses #76, #77, #78, #79, #80, #81, #82, #83, #84, #85, #89, #90, #91 and #97. Implementation issues are closed individually with final verification explicitly deferred at the user's request. #48 remains deferred.

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

Fix CookieMonster V2 session retention, desktop permissions and activation, panel navigation, refresh failures, browser request limits, timeline scrolling, portable SDK skill discovery and early mobile drawer dismissal. Closed-session transcripts are released, permissions survive multiple main windows, activation restores the main UI independently of hidden workers, and open-with requests enforce the native application allowlist.

Give CookieMonster its own Linux executable, launcher, icon and metainfo so it can coexist with OpenCode. Integrate the optional private WPP backend at 8bb2cd4e7d0588c3ef37735453830e7206f5a455. WPP access uses an automatically managed per-user credential. Browser requests have a 64 KiB UTF-8 admission limit and retain native permission grants and the approved trust of the authenticated backend across tasks.

The private backend was merged in https://github.com/trmosala/CM-WPP-Bits/pull/28. Its 21 verified issues are closed; genuine signed-in WPP capture remains open as private issue #16. The legacy DOM assertion and viewport-order fix is published separately and is not merged into V2.

How did you verify your code works?

Final source is 68ab9205709350385471eb2d383b3cff3c18e678. The user requested completion and closure of all implementations before one final test pass. Automatic CI is therefore deferred for this update; final platform qualification is outstanding. No cancelled job is counted as a completed pass.

Existing evidence from before testing was stopped:

  • All 630 private backend tests passed with 3,151 assertions. All 59 host contracts against the actual private Git pin passed with 657 assertions.
  • Canonical lint/type checks passed all 37 packages. Strict changed GUI lint reported zero problems across 55 files. A generic build passed with the private dependency physically removed. WPP Linux package/archive checks and disposable coexistence alongside official OpenCode 1.19.3 passed for the unchanged identity implementation.
  • Real Electron 44.4.5 fixtures covered multiple-window permissions and isolated out-of-process SSE/JSON-XHR capture. Those local native checks used --no-sandbox, so OS sandbox qualification is not claimed.
  • The cancelled combined run recorded 28 successful Linux package suites and all 449 Linux production browser cases before cancellation. Windows recorded 448 browser passes plus one SSH reconnection case that passed on retry. Component phases and Windows unit completion remain unqualified. The available Windows unit log is missing; its metadata says cancelled.
  • The Corvu drawer fix for fix(app): release focus after immediate mobile drawer dismissal #97 has a normal-motion negative control that fails before the patch and normal/reduced-motion checks that pass after it. The final repeated local target run passed 16/16. Full component verification is still part of the deferred final pass.
  • The transcript benchmark has a disclosed, uncertain 57 ms increase in one warm resized stabilization median, about 25.7% from 20 samples. Its p95 was similar.

The last completed Mac qualification attempt passed signing, notarization, stapling, mounted-installer validation, initial main-window/lifecycle adoption and hidden-worker setup. It found zero native activation events because AppleScript activate did not send the Dock-style reopen event. The script now sends native reopen before activate, preserving the actual-event, visible-main-window and hidden-worker assertions. The subsequent run was cancelled at the user's request, so the corrected packaged activation remains unverified.

The CI cancellation correction changes three always() build steps to !cancelled(). They continue after failures but stop on explicit cancellation; artifact cleanup remains enabled.

Genuine signed-in WPP capture remains unverified because the installed WPP window returned HTTP 401. The exact command and installation for historical kernel OOM victims were not retained, so correlation with the independently reproduced DOM assertion spike is not claimed as proof. The legacy fix remains published separately on bounded-dom-tests and is not ported into V2.

No release was published. The installed CM and backend were not updated or restarted. Issue closure records will distinguish implementation completion from deferred verification.

Screenshots / recordings

No new recording is attached. Verification currently uses automated browser geometry checks, real Electron fixtures and the disposable packaged Linux renderer. Final platform qualification remains pending as described above.

Checklist

  • I have tested my changes locally
  • I have not included unrelated changes in this PR

tiisetso added 30 commits October 20, 2026 09:02
The client data cache kept every session's transcript and pending rows
after its tab closed, and late reads, older pages and later events
could bring evicted rows back.

createSessionRetention, wired in createServerController, evicts a
session family once no open tab, session route, running work or
attention request has kept it for one second with no navigation
pending. Rows that return after eviction are released at the next
deadline. Eviction keeps unacknowledged submissions and sessions still
being created, so the policy retries until they settle.

Adds data.session.cached, which lists sessions holding transcript or
pending rows and notifies only when that set changes.

Refs #77
Replace unavailable Blacksmith labels with ubuntu-24.04 and windows-2025
for PR tests and duplicate checks. Preserve every job, matrix entry and
command while allowing this fork's queued checks to start.
The virtualizer core's reset-delay fallback ends a scroll with the offset read by the last scroll event. When a history prepend or anchor write moves scrollTop and a long task holds that write's own scroll event past the delay, the stale end delivery resets the core offset. Later prepend compensation then targets the stale offset, so loaded history rows are skipped. End-of-scroll deliveries now report the live scrollTop; scrolling deliveries are unchanged.

Refs #89
Pin Chromium's reduced-motion preference so SVG images animate on Windows runners, and give each compared fixture row its own layer so identical rows raster from the same tile origin. Assertions, tolerances and timings are unchanged.

Refs #90
@trmosala
trmosala marked this pull request as ready for review October 21, 2026 02:23
@trmosala
trmosala merged commit 290d7d6 into CM-Split-v2 Oct 21, 2026
2 checks passed
@trmosala
trmosala deleted the v2-verified-fixes branch October 21, 2026 02:23
@trmosala

Copy link
Copy Markdown
Owner Author

The one final verification pass ran after all 36 scoped implementation tickets were confirmed closed.

Source: public PR #96, merge 290d7d6353c0c7a4a356af780b5687f2dcfe5cfc, private pin 8bb2cd4e7d0588c3ef37735453830e7206f5a455.

  • Public run 39533186971 completed with an overall failure due to one Windows unit fixture. Linux unit/package/SDK/compiled-service/Node/generated checks all passed. Linux production browser 449, service worker 12, session UI 135 and app components 121 all passed.
  • Windows browser/component phases passed with automatic retries: production 445 first-pass plus four flaky cases, service worker 12 passed, session UI 134 first-pass plus one flaky case, app components 120 first-pass plus one flaky RTL overflow-layout case. The new controlled early-dismissal regressions passed. These are six intermittent cases, not a clean first-attempt Windows result.
  • The only failed unit test was Session.shell > runs shells concurrently with an active model and waits for each shell's own completion, at its inner five-second process-start file gate. The remaining Windows compiled service and Node build gates passed. Windows app and CLI unit packages did not complete after Turbo stopped on the core failure, so no Windows pass is claimed for those packages.
  • PR #98, merged as ac02f1195eb5df4fef02b35d30ecb910e5a3d374, adjusts that Windows startup gate to 15 seconds and the enclosing Windows fixture to 30 seconds. Other-platform deadlines and all concurrency/completion assertions remain. It changes no production source. Canonical lint/type checks passed 37/37, but the adjustment was not retested, honoring the single-pass instruction. No further test run was dispatched.
  • Mac run 39533190446 passed signing/notarization/stapling/mounted-installer checks and real native activation with a restored visible main window and preserved hidden worker. It also passed all 630 private tests with 3,151 assertions and all 59 pinned host contracts with 657 assertions. Release publication was skipped.

Genuine signed-in WPP capture remains unverified; the installed window returns HTTP 401 and a readiness question is pending. OS sandbox qualification of that capture is not claimed. Exact historical OOM-victim command/installation attribution remains unavailable, as recorded earlier.

All 36 scoped implementation issues remain closed; #48 remains explicitly deferred and open. The installed CM/backend were not restarted or updated. The memory watcher remains active.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant