Skip to content
This repository was archived by the owner on Oct 14, 2026. It is now read-only.

ci: sanity checkout fetch-depth 2 so parent-pin tests can resolve HEAD^ - #157

Merged
cursor[bot] merged 1 commit into
mainfrom
cursor/ci-sanity-fetch-depth-ff72
Sep 27, 2026
Merged

cursor[bot] merged 1 commit into
mainfrom
cursor/ci-sanity-fetch-depth-ff72

Conversation

@d6g8k5htny-coder

@d6g8k5htny-coder d6g8k5htny-coder commented Sep 27, 2026 •

Copy link
Copy Markdown
Owner

Scientific effect: NONE. One-line CI checkout change; no scripts, registers, claims, premises, prizes or lemma state touched.

Why

sanity is red on every main push since Batch 832. The two failures on the latest completed main run (36350109239, b9a1475f) are:

FAILED tests/test_intent.py::test_batch832_tip_or_eng_parent_pin_repair
FAILED tests/test_intent.py::test_batch833_tip_or_eng_parent_pin_repair
  subprocess.CalledProcessError: Command '['git', 'rev-parse', 'HEAD^']' returned non-zero exit status 128.
  fatal: ambiguous argument 'HEAD^': unknown revision or path not in the working tree.

Both Batch 832/833 parent-pin tests call git rev-parse HEAD^. The sanity job uses actions/checkout@v4 with the default fetch-depth: 1, so the checkout has no parent commit and the tests error before reaching their assertions. The loop runs the suite locally against a full clone, so it does not see this.

Change

.github/workflows/ci.yml, sanity job only: fetch-depth: 2. The four gate jobs (portable-patches-on-main, path-c-applied-bundle-dry-apply, land-workflows-dry-run, research-stack-status-guard) are untouched, so this does not overlap #155's gate rewiring (#155 adds a step after Install pytest; this touches only the checkout step above it — clean merge).

Evidence

  • Local full tests/ suite on this branch in a full-history clone: exactly one failure, test_batch833… at tests/test_intent.py:72485 (trial.tip_sha == repair.trial_tip_after, the loop's own pin, already advanced by Batch 834/835).
  • CI run 36350681045 on this branch: four gates green; test_batch832… passes (was HEAD^ error on main); test_batch833… reaches its assertion and fails on the pin (f7db5940… != 78ebd3b0…); one unrelated timing flake test_batch267_when_writable_dual_daemon_status_race (passes on main, on Dropbox intake digest — Dylan shared math corpus pointers #149, and locally).

What this does and does not fix

  • On main pushes, HEAD^ now resolves, so the 832/833 tests reach their real assertions.
  • The Batch 833 pin is the loop's domain (same shape it relaxed for 832 with "Batch 833+: inv tip_sha advances with parent_pin"); this PR only removes the shallow-checkout error class.
  • On pull_request runs HEAD is the synthetic merge commit and HEAD^ is the base tip, so parent-pin equality checks are still expected to differ on PR branches (as on Dropbox intake digest — Dylan shared math corpus pointers #149 run 36349927928); that is not a property of any PR's payload.
Open in Web Open in Cursor 

Co-authored-by: Dylan Roy <d6g8k5htny@privaterelay.appleid.com>
@cursor

cursor Bot commented Sep 27, 2026

Copy link
Copy Markdown
Contributor

Evidence (run ff72) — scientific effect NONE

Full tests/ suite run locally on this branch (origin/main f7db5940 + the one-line change) in a full-history clone, so HEAD^ resolves as it will in CI after this lands:

FAILED tests/test_intent.py::test_batch833_tip_or_eng_parent_pin_repair  (tests/test_intent.py:72485)

That single remaining failure is assert trial.get("tip_sha") == repair.get("trial_tip_after") — the Batch 833 pin that the loop's Batch 834/835 pulses already moved past (inventory details[trial].tip_sha is now 3d22b453…, the pin is 78ebd3b). It is the same shape the loop relaxed for 832 ("Batch 833+: inv tip_sha advances with parent_pin"), so it is expected to clear on the loop's next test touch. test_batch832_tip_or_eng_parent_pin_repair passes once HEAD^ exists.

Against main's depth-1 CI (run 36350109239) both 832 and 833 die earlier with fatal: ambiguous argument 'HEAD^', so with this PR sanity on main drops from two error-class failures to at most one loop-pin failure. #155's ci.yml change adds a step after Install pytest; this PR only touches the checkout step above it, so the two merge cleanly.

@cursor

cursor Bot commented Sep 27, 2026

Copy link
Copy Markdown
Contributor

CI result for this branch (run 36350681045) — scientific effect NONE

Four gate jobs green. sanity:

  • test_batch832_tip_or_eng_parent_pin_repair — now passes (was fatal: ambiguous argument 'HEAD^' on main).
  • test_batch833_tip_or_eng_parent_pin_repair — reaches its assertion and fails on the pin as predicted: trial.tip_sha 'f7db5940…' == repair.trial_tip_after '78ebd3b0…'. Loop pin debt, not checkout depth.
  • test_batch267_when_writable_dual_daemon_status_race — FileNotFoundError: …/ww.status.json; the daemon had not written its status file when the test read it. This is a timing flake (the same test passed on main run 36350109239, on Dropbox intake digest — Dylan shared math corpus pointers #149's run, and locally on this branch); unrelated to this change.

So the shallow-checkout error class is gone; what remains on sanity is the loop's own 833 pin plus one known-flaky race test.

@cursor
cursor Bot merged commit 3d3e1b8 into main Sep 27, 2026
4 of 5 checks passed
cursor Bot pushed a commit that referenced this pull request Sep 27, 2026
…-ci-deduplicate-premerge-20260926 to re-run sanity

Co-authored-by: Dylan Roy <d6g8k5htny@privaterelay.appleid.com>
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants