ci(gate): raise the default full-gate timeout to 90m fleet-wide - #45
Merged
Conversation
This was referenced Aug 18, 2026
Merged
forkwright
added a commit
to forkwright/hamma
that referenced
this pull request
Aug 18, 2026
…99) The pinned reusable SHA (df92942b, 2026-08-15) predates the 90m full-gate-timeout default raise (forkwright/.github#45), so this repo keeps the stale 45m budget forever — the cold-cache cancel class measured today (logismos died at 45m30s mid apt-install). Re-pin to d74efb1d2d87 (current .github main, which carries the 90m default + the release-please Bot-author waiver fix). The stale-pin class is kanon#3515 (.github has no tags, so dependabot cannot propose these bumps — nothing reports the drift). Co-authored-by: forkwright <cody@forkwright.com>
forkwright
added a commit
to forkwright/sphragis
that referenced
this pull request
Aug 18, 2026
…42) The pinned reusable SHA (df92942b, 2026-08-15) predates the 90m full-gate-timeout default raise (forkwright/.github#45), so this repo keeps the stale 45m budget forever — the cold-cache cancel class measured today (logismos died at 45m30s mid apt-install). Re-pin to d74efb1d2d87 (current .github main, which carries the 90m default + the release-please Bot-author waiver fix). The stale-pin class is kanon#3515 (.github has no tags, so dependabot cannot propose these bumps — nothing reports the drift). Co-authored-by: forkwright <cody@forkwright.com>
forkwright
added a commit
to forkwright/heurema
that referenced
this pull request
Aug 18, 2026
…31) The pinned reusable SHA (df92942b, 2026-08-15) predates the 90m full-gate-timeout default raise (forkwright/.github#45), so this repo keeps the stale 45m budget forever — the cold-cache cancel class measured today (logismos died at 45m30s mid apt-install). Re-pin to d74efb1d2d87 (current .github main, which carries the 90m default + the release-please Bot-author waiver fix). The stale-pin class is kanon#3515 (.github has no tags, so dependabot cannot propose these bumps — nothing reports the drift). Co-authored-by: forkwright <cody@forkwright.com>
forkwright
added a commit
to forkwright/epistole
that referenced
this pull request
Aug 18, 2026
…123) The pinned reusable SHA (df92942b, 2026-08-15) predates the 90m full-gate-timeout default raise (forkwright/.github#45), so this repo keeps the stale 45m budget forever — the cold-cache cancel class measured today (logismos died at 45m30s mid apt-install). Re-pin to d74efb1d2d87 (current .github main, which carries the 90m default + the release-please Bot-author waiver fix). The stale-pin class is kanon#3515 (.github has no tags, so dependabot cannot propose these bumps — nothing reports the drift). Co-authored-by: forkwright <cody@forkwright.com>
forkwright
added a commit
to forkwright/epitelesis
that referenced
this pull request
Aug 18, 2026
The local hybrid-gate adapter hardcoded a 45m timeout on full-gate-build — the one consumer the reusables 90m default raise (forkwright/.github#45) cannot reach. Same cold-cache cancel class measured across five repos today (kanon#10081). Ceiling for runaway jobs, not a target. Co-authored-by: forkwright <cody@forkwright.com>
forkwright
added a commit
to forkwright/thumos
that referenced
this pull request
Aug 18, 2026
…efault (#853) Two budget fixes from the kanon#10081 fleet audit: 1. kernel (i686 tests + armv7a build): warm p95 is 26m of its 30m budget across six sampled runs — a cold cache cancels it. Raised to 60m. 2. The pinned hybrid-gate SHA (230c932f) predates the 90m full-gate default (forkwright/.github#45), so thumos kept the stale 45m budget. Re-pinned to d74efb1d2d87 (current .github main; also carries the release-please Bot-author waiver fix). Co-authored-by: forkwright <cody@forkwright.com>
forkwright
added a commit
to forkwright/epitelesis
that referenced
this pull request
Sep 3, 2026
…tion (#36) ## Finding Two false assertions in this repo's own documentation: 1. `_llm/current_state.toml`'s `contract_status` claims the breaking v1 invocation/supervisor contract is "implemented on the convergence branch and awaits exact-head hosted compilation and gate evidence." No convergence branch exists (`gh api repos/forkwright/epitelesis/branches` lists only `main`), and the contract has been shipped and enforced since v0.3.0. 2. `hybrid-gate-adapter.yml`'s `timeout-minutes: 90` comment cites a kanon issue number that does not exist in that repo today. The number it actually meant is `forkwright/.github#45`, a merged PR that raised the fleet reusable's `full_gate_timeout_minutes` default to 90. ## Evidence - `gh api repos/forkwright/epitelesis/branches --jq '.[].name'` → `main` only. - `CHANGELOG.md` v0.3.0 entry (2026-07-28): "⚠ BREAKING CHANGES: enforce the v1 invocation lifecycle contract (#17)". - `gh pr view 34 --repo forkwright/epitelesis --json statusCheckRollup` and the same for PR #35: both show `gate / full-gate-build` and `gate / gate-attestation` green, merged onto `main` on 2026-08-25/26. - `gh issue list --repo forkwright/kanon --state all --limit 1` → kanon's highest issue number is far below the one cited in the adapter comment; the cited number does not resolve (`gh issue view` on it 404s). - `gh pr view 45 --repo forkwright/.github --json title,state,mergedAt` → merged, "ci(gate): raise the default full-gate timeout to 90m fleet-wide" — this is what the comment actually meant to cite. ## Why this matters `current_state.toml` is declared as this repo's canonical machine-readable state doc (`_llm/current_state.toml`, listed in its own `source_docs`). An agent or tool reading it for the contract's real status is told to look for a branch that doesn't exist and to wait for evidence that already landed five weeks ago — the opposite of what the file exists to prevent. The dead issue citation in the workflow comment sends the next reader chasing a number that resolves to nothing, right at the point where the comment is explaining *why* a value needs manual upkeep at all (this job is a hand-copy of the fleet reusable, not a `workflow_call` consumer of it, so it never inherits the reusable's own default and must be kept in sync by hand -- that fact was missing from the original comment too). ## Desired correction - `current_state.toml`: state that the contract is implemented and enforced on `main`, verified by the hosted full gate, and refresh `updated`. Done in this PR. - `hybrid-gate-adapter.yml`: point the citation at the PR that actually raised the default (`forkwright/.github#45`), and state plainly that this file hand-copies the fleet reusable rather than calling it. Done in this PR. **Not done in this PR, reported separately:** the deeper substrate-drift this hand-copy represents -- replacing the ~185-line duplicate with a direct `uses: forkwright/.github/.github/workflows/hybrid-gate.yml@main` call, matching `forkwright/zetesis`'s pattern -- is out of scope here. Verified via `forkwright/zetesis`'s live branch protection (`gh api repos/forkwright/zetesis/branches/main/protection`), the fleet reusable's own aggregation job is named `gate`, so a direct call would produce required-check context `gate / gate`, not this repo's current `gate / gate-attestation` (confirmed via this repo's own branch protection). That rename would break branch protection until coordinated with a protection update, which is out of this lane's authority. Flagged inline in the workflow comment and reported to the orchestrator. Done when: `current_state.toml` and the workflow comment both state only claims that are true today, and neither cites a resource that doesn't exist. Co-authored-by: forkwright <cody@forkwright.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The 45m default canceled full-gate builds at 45m+seconds on five repos in one day (2026-08-18, cold caches after the billing-outage expiry): harmonia#718 (45m15s), stathmos#127 (30m15s, local gate), typikon#176 (45m20s), akroasis#392 (45m20s), aletheia#6888 (45m15s). Each repo bumped its own caller to 60-90m today; the default should cover the fleet cold-cache p95 so the next repo does not rediscover it. Ceiling for runaway jobs, not a budget. Repos with explicit overrides are untouched.
This also stops the per-repo whack-a-mole this failure shape produced — the shared reusable is the right rung for the default.