Skip to content

Harden Windows GOA monitoring and report delivery - #29

Open
Tiyouw wants to merge 4 commits into
enowdev:mainfrom
Tiyouw:codex/upstream-windows-runtime
Open

Tiyouw wants to merge 4 commits into
enowdev:mainfrom
Tiyouw:codex/upstream-windows-runtime

Conversation

@Tiyouw

@Tiyouw Tiyouw commented Nov 4, 2026 •

Copy link
Copy Markdown
Contributor

What and why

This change hardens Windows GOA monitoring and report delivery:

  • Classify verified Codex runtime helpers using same-session ancestry, executable paths and process start times. Unknown processes remain watched; repeated alerts are throttled while audit records remain.
  • Use the executable artwork for the Windows tray and fallback notification identity.
  • Deliver worker reports through supported hooks or retain them in Tasks. Never paste reports or press Enter in the user's composer. Keep task/session identities during deduplication and cancelled handoffs.
  • Update stall regressions to observe Tasks rather than obsolete composer injection. Preserve repeated reminders, single worker-hook firing, state recovery, settlement, restart persistence and remote audit checks. Add a platform-independent Tasks-to-hook regression with draft preservation.

Validation and remaining gates

CI run 41886161893 completed successfully on head 00f1a4f75dc1b49a4da9ae0c1615191fc3dd7c84 (PR merge checkout into base 048b78e). All four jobs passed:

  • Linux and macOS: full go test -json -timeout 15m ./... and make lint (go vet ./... and staticcheck v0.8.2).
  • Both native Linux/macOS test artifacts confirm all seven relevant GOA stall regressions executed and passed without skips, including repeated reminders, hook/Tasks delivery and restart coverage. The end-to-end case passed in 29.37s on Linux and 29.06s on macOS.
  • TestLongSessionListsStayQuick passed the unchanged 16ms/frame budget: Linux 9.12ms by state / 5.29ms by project; macOS 13.36ms / 11.14ms. This passing sample does not diagnose the earlier macOS timing failure or complete a comparable baseline/head/candidate timing investigation. The unverified sidebar optimization is not included.
  • Windows: the workflow's targeted icon/iconlib, agent-command, diagnostics and daemon command tests passed. This is not full Windows test/lint coverage or native shell-stall coverage; earlier local Windows shell-test skips are not counted as passes.
  • Cloud-postgres: the actual PostgreSQL cloud test job passed.

Local focused daemon/protocol tests and go vet ./daemon ./proto also passed. SGR mouse input has not been reproduced or fixed. Live local/remote and GUI validation remain unperformed; no rollout is included. Commits are DCO signed off. Ready for review does not imply merge or installation approval.

  • Linux/macOS full test and lint CI pass on the candidate
  • Seven native Linux/macOS stall regressions pass without skips
  • Long-session-list timing passes the unchanged budget in this CI run
  • Comparable macOS timing investigation beyond this passing run
  • Full Windows test/lint coverage
  • Live local and remote validation
  • Docs updated where behavior changed
  • Commits signed off

Signed-off-by: Ahimsa Jenar Bramsaifstyo Kusuma <tio.rmx@gmail.com>
Signed-off-by: Ahimsa Jenar Bramsaifstyo Kusuma <tio.rmx@gmail.com>
Signed-off-by: Ahimsa Jenar Bramsaifstyo Kusuma <tio.rmx@gmail.com>
Signed-off-by: Ahimsa Jenar Bramsaifstyo Kusuma <tio.rmx@gmail.com>
@Tiyouw
Tiyouw marked this pull request as ready for review November 4, 2026 18:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant