Repository navigation
quest: plan backporting the moq-gst formula test fix to release - #8076
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
No actionable comments were generated in the recent review. WalkthroughThe changes document the Homebrew formula test timeout that affected moq-gst tags 0.4.21–0.4.23. They record the planned backport of the formula-template change from Severity of issue fixed: Medium Merge Risk:
|
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed head 04b39a0a589031fcd88868fda018837d34aee9b1 against base f2f9bc935ee81689a9e4aec424fa28b925a1c29e.
One low-priority finding: correct the command and script named for reproducing the failing Homebrew step (inline). No other actionable findings in this two-file plan change. The three cited tag runs do fail the macOS formula step and skip Release; the latest 0.4.23 job log confirms the five-minute timeout after the XPC errors.
Direction: start with the existing formula difference. At this head, .github/homebrew/Formula/moq-gst.rb.tmpl:46-54 already documents the headless Homebrew sandbox hang and checks dynamic linkage instead, while the workflow separately loads the plugin outside the sandbox. The 0.4.23 tag still invokes gst-inspect-1.0 inside the formula test. Preserve both kinds of coverage when validating the release-line fix.
Limits: GitHub-only static review of the plan, workflow, installer and formula, plus existing job results/logs. I did not reproduce the macOS behavior, run quest checks or tests, or independently verify registry publication. The PR's Check workflow reports success. COMMENT only, not approval or a merge-readiness assertion.
There was a problem hiding this comment.
Important
The plan treats the brew test hang as unsolved, but main settled the same failure in #5841 and release never received that commit. Details below.
Reviewed changes
- New quest plan (
quest/m0/ci/gst-release-macos.md) — records the three missing moq-gst GitHub releases and plans the macOSbrew testfix. - m0 CI index (
quest/m0/ci/README.md) — links the quest under## Required.
The cited tag runs, the skipped Release job, the missing 0.4.21-0.4.23 releases, the published crates and Nightly's passing macOS job all check out, and quest check passes (596 documents). The existing inline thread on the command and script names is correct and still applies.
⚠️ The plan re-diagnoses a cause main already settled in #5841
#5841 (2026-10-17) found gst-inspect-1.0 hangs at startup inside Homebrew's test sandbox (GStreamer 1.28.8, macOS image 20261013) and chose to drop the call from the formula's test do, asserting the plugin exists and links libgstreamer-1.0 instead. release never received that change: the Test Homebrew formula against tarball step and the pre-#5841 formula arrived in the backport be7f785b at 2026-11-03 00:53Z, about an hour before 0.4.21 was tagged, while 0.4.20's macOS job had no such step. The remaining work is porting #5841's formula to release and cutting the next release, so the Plan should point there rather than ask for a fresh macOS reproduction. The Goal's "without ... skipping the test" also contradicts that settled fix and needs rewording to allow it.
Technical details
# Port the settled formula fix instead of re-diagnosing
## Affected sites
- `quest/m0/ci/gst-release-macos.md:25-28` — treats the cause as open and asks for a fresh macOS reproduction.
- `quest/m0/ci/gst-release-macos.md:7-8` — the Goal forbids "skipping the test", which is what the settled fix does.
- `.github/homebrew/Formula/moq-gst.rb.tmpl:45-53` — on `main`, `test do` no longer runs `gst-inspect-1.0`; it asserts the plugin exists and links `libgstreamer-1.0`.
- The `release` copy of the same file still runs `gst-inspect-1.0 moq` (last changed by the backport `be7f785b`).
## Required outcome
- The Plan names #5841's decision as the known cause and fix, and the remaining work as backporting that formula to `release` plus cutting the next release.
- The Goal permits the accepted fix instead of forbidding it.
## Evidence
- #5841 (merged 2026-10-17), decision 1: "gst-inspect-1.0 hangs at startup inside Homebrew's test sandbox ... A: drop the gst-inspect-1.0 call from the formula's test do and assert the plugin exists and links libgstreamer-1.0."
- `be7f785b` (release backport of #5216) committed 2026-11-03 00:53Z added the `Test Homebrew formula against tarball` step and the pre-#5841 formula; 0.4.21 was tagged 2026-11-03 02:11Z.
- 0.4.20's macOS job (123997434413) step list has no `Test Homebrew formula against tarball` step, which is why it passed.
- 0.4.23 job 128111899383: `gst-inspect-1.0` logs the LaunchServices/XPC errors at 19:47:17Z and Homebrew times out at 19:52:17Z.deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) | 𝕏
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
- Quest rewritten as a backport — the Goal now requires
releaseto carrymain's Homebrew formula test rather than forbidding the settled fix. - Settled cause recorded — the Plan names #5841 as the fix already on
main,be7f785bas thereleasebackport that carried the pre-#5841 formula, and the workflow smoke test as the plugin-load coverage that remains outside the sandbox. - Plan narrowed — the fresh macOS reproduction step is gone; the remaining work is the template backport and the next tag.
- Index bullet reworded —
quest/m0/ci/README.mdnow describes the backport.
I re-checked the new claims against the tree: origin/release's test do still runs gst-inspect-1.0 moq and was last touched by be7f785b17, which is contained only in origin/release; the release/main moq-gst.yml difference is limited to the Nix action and artifact pins; just gh gst-formula → sh/gh/gst-formula.py matches the workflow step; and moq-gst-v0.4.20 is still the latest published moq-gst release. The prior body finding is fully addressed.
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) | 𝕏
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Follow-up review of head f1707d59f2e80dfc969e7e2fd5128db61a57efb4 against base 8c61297f349794a780d346e3aee6f0ba90214513, compared with my previous review at 04b39a0a589031fcd88868fda018837d34aee9b1 and accounting for the intervening main merge.
No new actionable findings. My prior P3 is fixed: the plan now names just gh gst-formula "$VERSION" "$TARGET" dist and sh/gh/gst-formula.py. The substantive rewrite correctly narrows the work to backporting #5841's formula change. I verified that PR's merged fix, the pre-fix formula introduced by be7f785b17d24660d4a06f90ee657a62b522a53d, and the current release formula. The workflow's remaining main/release differences are the Nix action and artifact pins, as the revised plan says.
Direction: keep the backport limited to the formula test, retain the separate unsandboxed plugin-load smoke step, and require the next tag's successful Release job before calling the quest complete.
Limits: GitHub-only static review; no local quest checks, tests, macOS reproduction, backport execution, or publication performed. Check was still running when inspected. COMMENT only, not approval or a merge-readiness assertion.

Summary
Plans
quest/m0/ci/gst-release-macos.md(XS). Every moq-gst tag since 0.4.20 fails itsBuild tarball (aarch64-apple-darwin)job, becausebrew test moq-dev/ci-gst/moq-gsthangs for 5 minutes aftergst-inspect-1.0logs macOS XPC "Connection invalid" errors. TheReleasejob is skipped as a result, so the 0.4.21, 0.4.22 and 0.4.23 GitHub releases never published. Found while re-running moq-dev/moq#7936's failed tag runs for moq-dev/moq#7201; run 42129723974 failed the same way on its re-run. Nightly's same job passes onmain, because moq-dev/moq#5841 already replaced the formula's sandboxedgst-inspect-1.0call there andreleasenever got it (found by Pullfrog). The quest is therefore a backport of that template change.Decisions
Is "moq-gst releases publish again" (find why
brew testhangs on release tags, fix it at the root, publish the next release) the right goal?Milestone:
Public API: none. Wire: none.
Review
Codex
quest review, gpt-6.1-sol, medium. Round 2 reviewed head f1707d5 against origin/main (8c61297): 0 findings. Round 1 (04b39a0): 0 findings.Online review on 04b39a0: the P3 on reproduction commands is fixed (the plan now names
just gh gst-formulaandsh/gh/gst-formula.py). Pullfrog's finding that #5841 already fixed this onmainis adopted: the quest is rewritten as an XS backport.quest checkpassed (606 documents ok).(Written by Claude Opus 5.5 with high)
🤖 Generated with Claude Code