Skip to content

Bound Docker metadata queries when the CLI never calls back - #843

Merged
antropos17 merged 1 commit into
masterfrom
codex/docker-owned-timeout-20261011
Oct 10, 2026
Merged

antropos17 merged 1 commit into
masterfrom
codex/docker-owned-timeout-20261011

Conversation

@antropos17

Copy link
Copy Markdown
Owner

Behavior

When Docker's metadata CLI does not invoke its callback after its own timeout, refresh() previously remained pending and all later refreshes coalesced onto it. Add an independent three-second watchdog that settles the observation as unavailable, preserves dated candidates, prevents endpoint fallback on timeout, and allows the next eligible refresh. The watchdog, execFile timeout and shutdown all send SIGKILL only to the owned child; late callbacks cannot overwrite a newer observation.

Verification

  • The affected Docker scanner suite passed 60/60 cases, including a never-callback child, stale candidate retention, Windows endpoint selection, late callback, cadence recovery, spawn failure, stop and throwing kill.
  • Targeted ESLint, Prettier and git diff --check passed.
  • Three independent GPT-6.1 Sol reviews passed the frozen tree cd59fe7f04418d00eaf61faa7c8bbfa51f86380b; a reviewer found and verified the SIGKILL correction before commit.
  • The five required hosted CI contexts will validate this PR head.

This is a bounded issue-#8 reliability fix. No live Docker daemon, WSL or Windows VM qualification was performed, and the full issue remains open.

@antropos17
antropos17 merged commit 7e572b9 into master Oct 10, 2026
5 checks passed
@antropos17
antropos17 deleted the codex/docker-owned-timeout-20261011 branch October 10, 2026 23:14
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