Skip to content

Lower the esp-matter timeout to 60 minutes - #17

Merged
hacker-cb merged 1 commit into
masterfrom
lower-esp-matter-timeout
Aug 5, 2026
Merged

Lower the esp-matter timeout to 60 minutes#17
hacker-cb merged 1 commit into
masterfrom
lower-esp-matter-timeout

Conversation

@hacker-cb

Copy link
Copy Markdown
Contributor

180 was set against a QEMU leg that took 337 minutes, before any native leg had
been measured. It has been now — six legs across two runs:

amd64 arm64
Matter 1.6 10m06s 10m01s
Matter 1.5.1 9m34s 10m30s
Matter 1.4.2 9m03s 8m50s

Worst observed anywhere, including the pull-request run: 10m58s. So 60 keeps a
five- to sevenfold margin.

Deliberately not tightened further toward the observed number, because the failure
mode is quiet rather than loud: a killed leg takes its manifest job with it, the
other platform's image stays in GHCR unreferenced, and the published tags sit on
the previous build while the run reports failure. A ceiling that is too low costs
much more than one that is never reached.

CLAUDE.md now records the measurement instead of the plan to take it, and says
what to do before changing the number again. It also drops the sentence claiming
the manifest timeout lives in versions.json — it does not; only build jobs read
timeout_minutes, which was itself a finding on the previous PR.

Verification

  • ./scripts/check-versions.sh green
  • generated matrix carries the new value: idf-v5.5.5-matter-v1.6: timeout_minutes=60
  • codex-review (xhigh) — no findings; confirms the value propagates through the
    generated matrix and the documentation matches the configuration

This run rebuilds everything (images/versions.json is in the paths filter), which
also exercises the new ceiling against the real ~10-minute legs.

180 was set against a QEMU leg that took 337 minutes, before any native leg had
been measured. Six native legs across two runs now put the range at 8m50s-10m30s
on master and up to 10m58s on a pull request, so 60 keeps a five- to sevenfold
margin over the worst observed.

Left deliberately wide rather than tightened toward the observed number: the
failure mode is quiet. A killed leg takes its manifest job with it, the other
platform's image stays in GHCR unreferenced, and the published tags sit on the
previous build while the run reports failure - so the cost of a ceiling that is too
low is much higher than the cost of one that is never reached.

CLAUDE.md now records the measurement rather than the plan to take it, and says
what to do before changing the number again.
Copilot AI lite review requested due to automatic review settings August 5, 2026 09:06

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR updates the esp-matter image family’s CI build timeout to better reflect current native build performance, and updates the repo’s operational documentation to match the new measured baseline.

Changes:

  • Lower esp-matter’s timeout_minutes from 180 to 60 in images/versions.json.
  • Update CLAUDE.md to reflect measured native esp-matter build durations and clarify which jobs are (and are not) controlled by timeout_minutes.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File Description
images/versions.json Reduces esp-matter build timeout to 60 minutes to align CI ceiling with measured native runtimes.
CLAUDE.md Refreshes CI timeout documentation to reflect the new measurement-driven rationale and current configuration behavior.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@hacker-cb
hacker-cb merged commit bf18308 into master Aug 5, 2026
18 checks passed
@hacker-cb
hacker-cb deleted the lower-esp-matter-timeout branch August 5, 2026 09:21
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.

2 participants