Skip to content

Commit 5c9bfc4

Browse files
luisleo526claude
andcommitted
docs: bring the CI pages in line with the smoke-leg post-merge workflow
The change that moved the full profiles out of GitHub CI left sentences in docs/ci.md and docs/reference/ci-runner-routing.md that still described the old post-merge workflow. They now say what runs: smoke legs in ci-post.yml, the full profiles and the full parity in the lab's verify, no preflight-slow job, and the measured time of the parity subset job (6.8 minutes in run 43308686886, against 13.5 and 14.3 before its strategy libraries built in parallel). docs/ci.md keeps its line count. Two comments in scripts/test_ci_preflight.py call a CI leg a leg. Doc reverts, each replaced by a sentence about the new shape: - "`preflight-slow` runs exactly those stages with `--tier slow`" and "Toolchain setup and the hash-pinned pytest install stay in the slow job." (no GitHub job runs the slow tier; the lab's preflight step runs every tier). - "the caller's `build` aggregate still requires the reusable workflow to succeed" (ci-post.yml no longer calls native-live.yml). - "retains the subset through `corpus-parity.yml` (`mode: subset`) alongside its full-parity call" (there is no full-parity call). - "The GKE Linux arm64 Debug leg runs after merge; a red result there blocks the next release cut under the same fix-or-revert rule." and "The GKE Linux arm64 Debug leg also runs before merge when `arm_relevant=true`; a red post-merge result blocks the next release cut under the same rule." (the leg is a smoke now; the full arm64 Debug profile left GitHub CI). - "the full-profile commands in `ci-post.yml` carry no explicit `--exclude-label` arguments" and "PR CI keeps the Release smoke selected by CTest label `fast`." (the commands run in the lab's verify; GitHub CI keeps the smoke on every configuration). - "Pull requests run the Ubuntu amd64 Release smoke; post-merge runs Ubuntu amd64 and arm64 Release/Debug." and "Both architectures use the same full-profile commands." (the `build`, `build-arm` row of the job table). - docs/reference/ci-runner-routing.md: "Post-merge `ci-post.yml` runs the full profiles on x86 and arm64, plus native and parity callers; there is no macOS cell." and "the post-merge legs (`ci-post.yml`) run their full profile populations, and a documentation-only PR runs admission, preflight and the docs build while skipping proof legs." Co-Authored-By: Claude Fable 5.2 <[email protected]>
1 parent 4ab0aa4 commit 5c9bfc4

3 files changed

Lines changed: 20 additions & 20 deletions

File tree

‎docs/ci.md‎

Lines changed: 16 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -104,9 +104,9 @@ classification selects the ordinary fast tier and fails the aggregate.
104104
also avoids the Doxygen build.
105105

106106
`preflight --tier fast` excludes `verifier-tests`, `kernel-residuals-tests`,
107-
`unittest-shards-tests` and `run-json-harness-tests`; `preflight-slow` runs
108-
exactly those stages with `--tier slow`. Toolchain setup and the hash-pinned
109-
pytest install stay in the slow job. A plain local invocation below retains
107+
`unittest-shards-tests` and `run-json-harness-tests`. No GitHub job runs
108+
`--tier slow` any more: the lab's `preflight` step runs every tier, those
109+
four stages included. A plain local invocation below retains
110110
all stages and its existing event-aware verifier-class selection. `runner-allowlist` independently scans workflow runner labels against the committed mirror in the fast tier's parallel group; `preflight-tests` runs every discovered method across up to eight shards, sized to available CPUs, with separate logs in `build-ci-preflight/preflight-tests-shards/`.
111111

112112
The smoke configures Release, builds `pineforge_pr_smoke` and directly runs
@@ -675,7 +675,7 @@ checks every runner compile unit for instrumentation, and stops on a race.
675675
Both matrices keep `fail-fast: false` and build pinned WebSocket curl cold. A failed cell does not cancel its peers.
676676
Native diagnostic artifacts carry the shard index; compiler and curl installations have no Actions cache.
677677
The `native-live (native)` aggregate requires all four successful summary proofs;
678-
`native-live-gate` requires it and both sanitizer cells, refusing missing or skipped work; the caller's `build` aggregate still requires the reusable workflow to succeed.
678+
`native-live-gate` requires it and both sanitizer cells, refusing missing or skipped work. `ci-post.yml` no longer calls this workflow, so no `build` aggregate waits for it.
679679
Only native shards accept the caller's existing PR-only slow-label exclusion; dispatch always runs the full set.
680680

681681
`native_live_equivalence_e2e_<probe>` names eight rows in `runner/CMakeLists.txt`,
@@ -1138,8 +1138,8 @@ judge <1 s — **94 s** end to end over an up-to-date build directory, against
11381138

11391139
`gke-pr` inlines the corpus subset for ordinary code PRs and skips it only
11401140
after the base classifier proves docs-only or GitHub-only. `ci-post.yml`
1141-
retains the subset through `corpus-parity.yml` (`mode: subset`) alongside its
1142-
full-parity call.
1141+
retains the subset through `corpus-parity.yml` (`mode: subset`) and makes no
1142+
full-parity call: the full parity is the lab's `parity` step.
11431143
The merge ruleset requires `pineforge/verify` and
11441144
`pineforge/parity`, posted by the maintainers' `lab verify` tooling after
11451145
full verification on the PR head's exact tree.
@@ -1540,8 +1540,8 @@ result on `main` in an excluded row or class (the Release and kernel-only
15401540
`slow` rows, or a SLOW verifier class) BLOCKS the next release cut, and the
15411541
author or the release owner fixes or reverts the change within the hour; the
15421542
red is not carried into a release.
1543-
The GKE Linux arm64 Debug leg runs after merge; a red result there blocks
1544-
the next release cut under the same fix-or-revert rule. The GKE Linux arm64 Debug leg also runs before merge when `arm_relevant=true`; a red post-merge result blocks the next release cut under the same rule.
1543+
The Linux arm64 Debug smoke runs after merge; a red result there blocks
1544+
the next release cut under the same fix-or-revert rule. The same smoke also runs before merge when `arm_relevant=true`. The full arm64 Debug profile no longer runs in GitHub CI (see "Where the full suites run").
15451545

15461546
### Documentation-only changes
15471547

@@ -1765,11 +1765,11 @@ reader, so a preceding job cannot authorize a GitHub-only skip. Full CI keeps
17651765
every proof dependency and both preflight tiers. Docs deployment and parity's
17661766
HTTPS submodule clone retain their existing credential handling behind authz.
17671767

1768-
**2026-10-24 clarification:** the full-profile commands in `ci-post.yml`
1769-
carry no explicit `--exclude-label` arguments. The verifier's Debug-type
1770-
default excludes `slow` from Debug and sanitizer runs on every event,
1768+
**2026-10-24 clarification, updated 2026-11-13:** the full-profile commands
1769+
(now run by the lab's verify; `ci-post.yml` keeps smoke legs only) carry no
1770+
explicit `--exclude-label` arguments. The verifier's Debug-type default excludes `slow` from Debug and sanitizer runs on every event,
17711771
including non-PR events; Release, kernel and native retain every row.
1772-
PR CI keeps the Release smoke selected by CTest label `fast`.
1772+
GitHub CI keeps the smoke selected by CTest label `fast`.
17731773

17741774
Baseline promotion uses hosted `ubuntu-24.04` for its GitHub App
17751775
secrets and inherited `id-token: write` permission, behind its `release-bot` environment
@@ -1867,15 +1867,15 @@ measurements on the new runners.
18671867
| Job | Platform maximum (min; no declared limit) | Measured basis |
18681868
| --- | --- | --- |
18691869
| `preflight` | 360 (was 45) | 7.3-9.8 min on the standard runner through 1a0e7ea1; cancelled at 10.3 min at 0d76a099, 593 s into `verifier-tests`, which takes about 15 min there at that tree (above), so preflight about 16 min. 4.8-8.2 min on the verification hosts. Preflight supplies no stage deadline -- the old 2400 s stage bound is gone (owner rule, 2026-10-12), and owner #827 (2026-10-20) removed every declared job timeout; the native-live aggregate job names a stop at GitHub's own 6 h platform maximum `HANG (re-run)` in its log; the check still fails. |
1870-
| `build`, `build-arm` | 360 (was 75) | Pull requests run the Ubuntu amd64 Release smoke; post-merge runs Ubuntu amd64 and arm64 Release/Debug. Both architectures use the same full-profile commands. The following historical measurements include retired macOS Debug cells on pull requests. Measured while Release ran its full population on every event, a fork's pull request included: 30.3 min Ubuntu and 32.0 min macOS on the standard runners at 0d76a099, where the full Debug legs took 43.5 and 37.0 min. Pull-request Debug sets then took 17.2-17.7 min. 11.1-18.4 min on the verification hosts. GKE Linux arm64 timings await measurement. |
1871-
| `sanitizers` | 360 (was 120; the verifier's CTest stage bound was 60 min full, 30 min PR) | A fork's pull-request set took 35.2 min on the standard runner, whose compile alone took 1277-1335 s (pull-request run 36139176292, and main at 0d76a099). Full runs on the standard runner took 34.0-69.4 min (at 0d76a099 30 min of build and ABI providers, then 39 min of CTest, 1768 s of it the serial `test_ci_verify` row), and 18.2-19.0 min on the verification hosts. |
1872-
| `kernel-only` | 360 (was 60) | Push and dispatch events run the whole kernel set; a pull request's kernel-only leg excludes the `slow` rows, and that pull-request time is to be measured on real PR runs. Measured while every event ran the whole kernel set: 7.9-22.3 min on the standard runner, 6.5-9.8 min on the verification hosts. The hosted kernel-only run of 2026-10-12 lost its job to the old 1800 s stage bound (exit 124). |
1870+
| `build`, `build-arm` | 360 (was 75) | Pull requests run the Ubuntu amd64 Release and Debug smokes and, when `arm_relevant`, the arm64 Debug smoke; post-merge runs the Release and Debug smokes on Ubuntu amd64 and arm64. The full-profile commands run in the lab's verify since 2026-11-13. The following historical measurements include retired macOS Debug cells on pull requests. Measured while Release ran its full population on every event, a fork's pull request included: 30.3 min Ubuntu and 32.0 min macOS on the standard runners at 0d76a099, where the full Debug legs took 43.5 and 37.0 min. Pull-request Debug sets then took 17.2-17.7 min. 11.1-18.4 min on the verification hosts. GKE Linux arm64 timings await measurement. |
1871+
| `sanitizers` | 360 (was 120; the verifier's CTest stage bound was 60 min full, 30 min PR) | Since 2026-11-13 the post-merge job is a smoke under ASan and UBSan, and the full profile is the lab's `sanitizers` step. The measurements below are of that full profile. A fork's pull-request set took 35.2 min on the standard runner, whose compile alone took 1277-1335 s (pull-request run 36139176292, and main at 0d76a099). Full runs on the standard runner took 34.0-69.4 min (at 0d76a099 30 min of build and ABI providers, then 39 min of CTest, 1768 s of it the serial `test_ci_verify` row), and 18.2-19.0 min on the verification hosts. |
1872+
| `kernel-only` | 360 (was 60) | No GitHub job since 2026-11-13: the kernel profile is the lab's `kernel` step. What follows describes the former job. Push and dispatch events run the whole kernel set; a pull request's kernel-only leg excludes the `slow` rows, and that pull-request time is to be measured on real PR runs. Measured while every event ran the whole kernel set: 7.9-22.3 min on the standard runner, 6.5-9.8 min on the verification hosts. The hosted kernel-only run of 2026-10-12 lost its job to the old 1800 s stage bound (exit 124). |
18731873
| `native-live (native)` | 360 (was 75; verification step 65) | Main runs 37398893123, 37461653250 and 37478400505 spent 35.40–40.65 min in native, including 25.32–29.53 min in full CTest. Their matching PRs took 2.39–5.67 min in native CTest with the existing slow exclusion. The longest measured `test_ci_verify` row was 1695 s there and 2178 s on a c5.4xlarge Spot host, cut off by a trial deadline. The verifier supplies no default row or stage deadline and the workflow supplies no step deadline: its old 3600 s row bound is what collided with the 2820 s stage deadline on main's push run 38217844683. These are observed GitHub and Spot timings; final-head cold proof remains required. |
18741874
| `native-live (live-sanitizers)` | 360 (was 45; verification step 35) | The same three main runs spent 15.10–16.33 min in this profile after native. Its old per-row 900 s and stage 1200 s deadlines are gone; the verifier supplies no replacement, and the platform's job maximum catches a hang. |
18751875
| `native-live (live-tsan)` | 360 (was 45; verification step 35) | The old shared 60-min job left only 9.20, 4.67 and 2.67 min for TSan before cancellation. Matching passing PR profiles took 10.87, 10.93 and 7.70 min. Its old 900/1200-s row/stage bounds are gone like live-sanitizers'; the verifier supplies no replacement. |
18761876
| `native-live-gate` | 360 (was 5) | Seconds; failure, cancellation and unexpected skip all fail the aggregate. |
18771877
| `corpus-parity` | 360 (was 120) | 16.3-26.2 min on the standard runner, 10.1-10.3 min on the verification hosts. |
1878-
| `corpus-parity-subset` | 360 (was 30) | 1.8-4.7 min on the standard runner, 2.4-3.8 min on the verification hosts. |
1878+
| `corpus-parity-subset` | 360 (was 30) | 1.8-4.7 min on the standard runner, 2.4-3.8 min on the verification hosts. Those are of a shorter probe list. With 54 probes the job ran 13.5 and 14.3 min in runs 43205609467 and 43242369382, and 6.8 min in run 43308686886 once its strategy libraries built in parallel. |
18791879
| `build` (aggregate) | 360 (was 5) | Seconds. |
18801880
| `gke-pr / result` | Platform maximum | Seconds after the second compare read and job-result checks. |
18811881

‎docs/reference/ci-runner-routing.md‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
The merge ruleset requires `pineforge/verify` and `pineforge/parity` commit
22
statuses on the PR head's exact tree. The maintainers post them from their own (private)
33
verification tooling after full verification on their x86-64 build hosts and a
4-
parity verdict; an outside contributor cannot post them. GitHub Actions is advisory: `ci.yml` retains PR admission; `gke-pr.yml` runs Release and Debug fast builds, conditional arm64 Debug, preflight and the parity subset. Post-merge `ci-post.yml` runs the full profiles on x86 and arm64, plus native and parity callers; there is no macOS cell.
4+
parity verdict; an outside contributor cannot post them. GitHub Actions is advisory: `ci.yml` retains PR admission; `gke-pr.yml` runs Release and Debug fast builds, conditional arm64 Debug, preflight and the parity subset. Post-merge `ci-post.yml` runs Release and Debug smoke legs on x86 and arm64, a sanitizer smoke and the parity subset; the full profiles run in the lab's verify (`docs/ci.md`, "Where the full suites run"); there is no macOS cell.
55
The x86 PR smoke builds in `gke-pr.yml` run the `fast`-labelled CTest rows (`ctest -L '^fast$'`); the post-merge legs (`ci-post.yml`) run
6-
their full profile populations, and a documentation-only PR runs admission, preflight and the docs build while skipping proof legs.
6+
the same `fast` rows on every configuration, and a documentation-only PR runs admission, preflight and the docs build while skipping proof legs.
77
The seven PR/post-merge verification and documentation workflows start with the canon `authz` job and every other job in them needs it; release uses its status gate and baseline promotion its `promote-guard`. Pull-request
88
proof workflows and post jobs use exact hosted pins: `ubuntu-24.04` for x86 and `ubuntu-24.04-arm` for the arm64 leg. Promotion, the ruleset probe and docs deployment also run on hosted `ubuntu-24.04`, and keep their `release`/`release-bot` environments. Release's `release`, `prebuilt-x86`, `upload` and `notify-hub` jobs use hosted `ubuntu-24.04`; `prebuilt-arm` retains `ubuntu-22.04-arm`. `release` and `notify-hub` keep their `release` environment. Hosted PR light jobs and `docs.yml`'s `build` retain `ubuntu-24.04`; post guards and aggregates use the same pin.
99
Hosted fork PRs receive no secrets, so they need no VM-only fork-refusal hook. Retired main/post VM labels remain refused. The retired `gke-*` scale sets and paid larger-runner labels are refused too. Group 14's mirror `.github/runner-group-allowlist.txt` and group 15's mirror `.github/runner-group-15-allowlist.txt` are empty for the current hosted workflows. The exact prior isolated and hosted-matrix release forms remain admitted for rollback; restoring a protected route requires its mirror entry and live group membership.

‎scripts/test_ci_preflight.py‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -8195,7 +8195,7 @@ def test_the_build_aggregate_passes_a_docs_only_skip_and_nothing_else(self):
81958195
self.assertIn('amd64 smoke: skipped', result.stdout)
81968196
self.assertIn('arm64 smoke: skipped', result.stdout)
81978197
# Native x86 must succeed on every event, a passing docs-only push included:
8198-
# every other lane is set as in that passing run, so only this result decides.
8198+
# every other leg is set as in that passing run, so only this result decides.
81998199
for outcome in ('failure', 'cancelled', 'skipped', ''):
82008200
with self.subTest(native_x86=outcome):
82018201
result = subprocess.run(
@@ -8350,7 +8350,7 @@ def test_github_only_aggregate_defers_only_full_event_lanes(self):
83508350
{'build': 'BUILD_RESULT', 'build-debug': 'DEBUG_RESULT',
83518351
'build-arm': 'ARM_RESULT',
83528352
'corpus-parity-subset': 'PARITY_RESULT'})
8353-
# Every CI round fits 15 minutes: ci-post's own lanes are the amd64 and arm64
8353+
# Every CI round fits 15 minutes: ci-post's own legs are the amd64 and arm64
83548354
# smoke matrices and the sanitizer smoke (preflight-slow, the leg aggregates,
83558355
# kernel-only, native-live and the full parity call left ci-post for the
83568356
# lab's verify and parity steps).

0 commit comments

Comments
 (0)