Repository navigation
Intel macOS runs on the release line, not on each pull request (§8.3, §9.4) - #535
Conversation
The repository has about one hosted macos-15-intel runner and the host's workload takes about three hours of it per run, so whole-gate pull requests queued six hours and more for it. ci-plan.sh now writes `intel`: a ready pull request into a release line runs the whole gate without the Intel construction, lanes and aggregate, unless its head is a release cut (cut/v*) or its base is main; pushes and requests run them. A push landing a tree whose pull request run passed every job but the skipped Intel jobs (Intel context once and skipped) runs honesty and the Intel jobs only. Such a run never counts as the whole gate for the plan, a dispatch, or release-gate.sh, whose four-context rule already refuses a skipped Intel context. RP-14 pins the new plan cases and the Intel jobs' `intel` condition; RP-18 plants a non-Intel run as release evidence and refuses it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013Asa3QVAsEzzaQPMzJnNz1
afflom
left a comment
There was a problem hiding this comment.
P1 — This plan conflicts with the standing project instruction requiring successful full acceptance on the final head before merge. ci-plan.sh:213–216 skips Intel for ordinary ready release PRs, and dispatch:255–266 makes the same choice. Thus this PR's own planned green result cannot satisfy that requirement. A later merge-push Intel-only run, even with equal trees, cannot replace full pre-merge acceptance or be aggregated into a whole gate.
Please retain the full pre-merge Intel path and align its RP-14/SPEC expectations while keeping the valid evidence-reuse and duplicate-work reductions. The proposed policy text and body's claimed request do not override the current explicit instruction; no waiver is given here. Existing implementing owner retains this correction, with no competing patch or CI controls.
Source credit: exact five-file delta preserves strict partial-evidence classification and the unchanged release gate's refusal of incomplete host contexts; RP-14/RP-18 add corresponding refusal plants. No additional source defect established in this bounded review. Current gate is incomplete and intentionally skips Intel; no tests or completed checks were repeated. Receipt pr535-835005-source-policy-review/receipt.json. Stopped #431/oldTC03 scope was not reviewed.
|
Agreed: this conflicts with the rule that the final head passes full acceptance before merge. Moved to draft, and it won't merge unless the owner waives that rule for Intel. Without the waiver, the work that remains is cutting the Intel minutes each gate needs, within full pre-merge acceptance. Generated by Claude Code |
|
The owner waived full pre-merge acceptance for Intel macOS only (project thread, 2026-11-14 19:59Z). Ordinary pull requests into a release line skip Intel. Line pushes, cuts and pull requests into Generated by Claude Code |
|
Updated after external landing: #535 merged at22:40:06 as acc0f39, parents0a9891ea +83500515; actual treeabafd7e5 equals the reviewed head. Its final proposal run43635042065 contains64successful jobs and3skipped Intel contexts. Preserve this partial evidence honestly; neither tree equality nor later Intel-only execution establishes full pre-merge acceptance. Owner6883385017 reports an Intel-only waiver in another project thread. The current maintenance automation still explicitly requires full final-head acceptance before merge and forbids incomplete gates; that instruction remains unresolved here. This review does not adopt the reported exception or grant retrospective approval. Please reconcile the authoritative instruction with the automation. Existing release history is preserved; no revert, tag rewrite, duplicate gate or CI-control action is requested. |
Brings in #535 (Intel macOS on the line). The only conflict was the Unreleased CHANGELOG list; both entries are kept, the Lean 4.32.2 pin first. No file #535 touches names the Lean pin. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013Asa3M8ULnzEQwLLngyBBX
Takes #535's Intel rules of SPEC.md §9.4 and its changelog entry beside the forward-merge text; the RP-14 sentence keeps both the push triggers of ci.yml and reproducibility.yml and the forward merge's. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013Asa3M8ULnzEQwLLngyBBX
Brings in #535 (Intel macOS on the line). The only conflict was the Unreleased CHANGELOG list; both entries are kept. #535 adds no register IDs, so the count stays 497. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013Asa3M8ULnzEQwLLngyBBX
Brings in #431, #535, #519 and #518's merge. No conflicts. #507 already carried #518's change and its regenerated outputs (d954d7b), so the combined 1.2 identity is unchanged; #431 only widens the snapshot nesting the canonical writer accepts and does not change any written bytes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013Asa3M8ULnzEQwLLngyBBX
…#535, #536) into the arrays branch The generated outputs that both sides rewrote take release/v0.4's copy, or are removed where the line no longer has them; regenerate.yml rewrites the whole set from this tree. Lean moves to 4.32.2 with #518, so the array authority follows the other Lean rows: LEAN-ARRAY-4-32-1 becomes LEAN-ARRAY-4-32-2, cited at f3b06c705e6c. The three files its statement rests on (src/include/lean/lean.h, src/Init/Prelude.lean, src/Init/Data/Array/Set.lean) are byte-identical at f054605aea4b and f3b06c705e6c, so the statement carries over unchanged. runtime-1.2/library.toml keeps the line's toolchain and the branch's Arrays module. LANGUAGE_1_2_SEMANTICS_ID is the merged tree's, as RP-10 recomputes it. CHANGELOG and VERIFICATION keep both sides' entries. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Requested by Alex · project thread
The org gets about one
macos-15-intelrunner, and a whole gate needs ~190 Intel runner-minutes (construction 19 + lanes 50/43/45/32, run 43515398695). With several pull requests gating at once, the Intel lanes queue 6–7 h (#405, #518, #520, #431, #526). More or fewer lanes would not change the Intel minutes.Plan (
ci-plan.sh, new outputintel):cut/v*head), a pull request intomain, a push, and a plain request.Release evidence is unchanged.
release-gate.shalready requires all four host contexts concludedsuccessin one run. The cut's own run includes Intel, sotag.ymlandrelease.ymlread a complete run. RP-18 now plants an Intel-skipped run as evidence and refuses it.Pins. RP-14 checks every plan decision against planted events, including defects such as a failed Intel job, another skipped job, a missing context, a missing host or a missing leg, so a non-Intel run can't count as the whole gate. RP-14 also pins the jobs'
intelconditions. SPEC §8.3 and §9.4 are updated, and CHANGELOG has an entry.Checked locally:
bash -n ci-plan.shcheck-test-inventory,validate-modelandvalidate-spec-linkscargo fmt --check-D warningson repo-conformance.github/scriptsunittestsThis pull request's own run skips Intel, by its own plan. Its merge push runs honesty and Intel.
Generated by Claude Code