Rewrites the note under the canonical Go make test example in prompts/REPO_POLICIES.md, for #77. The note named a cache-busting build argument that --no-cache replaced, and said Go's cache survived in earlier image layers.
The note now says where Go's test result cache can replay a pass: on a developer's machine, where the Makefile target runs, so -count=1 stays on both invocations there. The test phase of the Dockerfile has nothing to replay, because its base image holds no result for the repo's tests and no step before go test runs a test. The note says that phase needs no -count=1, because a reader comparing the two examples would otherwise wonder why only one carries it.
Not visible in the diff:
The issue guessed the pre-commit hook also runs go test on the host. It does not: the hook runs script/check, which builds the test phase in Docker, so the note does not mention the hook.
On this host's Docker (29.8), --no-cache also gives a BuildKit cache mount a fresh, empty directory, so even a phase that mounted GOCACHE would have nothing to replay under the scripts. The note leaves this out.
Judgement call: the first paragraph said the rerun "reproduces a failure instead of replaying it". Go stores only passing results, so it now says neither run can report a stored pass.
The link to issue 26 is gone; the note no longer refers to it.
Model: opus-5-5
Rewrites the note under the canonical Go `make test` example in `prompts/REPO_POLICIES.md`, for https://git.eeqj.de/sneak/prompts/issues/77. The note named a cache-busting build argument that `--no-cache` replaced, and said Go's cache survived in earlier image layers.
The note now says where Go's test result cache can replay a pass: on a developer's machine, where the Makefile target runs, so `-count=1` stays on both invocations there. The `test` phase of the `Dockerfile` has nothing to replay, because its base image holds no result for the repo's tests and no step before `go test` runs a test. The note says that phase needs no `-count=1`, because a reader comparing the two examples would otherwise wonder why only one carries it.
Not visible in the diff:
- The issue guessed the pre-commit hook also runs `go test` on the host. It does not: the hook runs `script/check`, which builds the `test` phase in Docker, so the note does not mention the hook.
- On this host's Docker (29.8), `--no-cache` also gives a BuildKit cache mount a fresh, empty directory, so even a phase that mounted `GOCACHE` would have nothing to replay under the scripts. The note leaves this out.
Judgement call: the first paragraph said the rerun "reproduces a failure instead of replaying it". Go stores only passing results, so it now says neither run can report a stored pass.
The link to issue 26 is gone; the note no longer refers to it.
Model: opus-5-5
Rebased onto current next: resolved the TODO.md Completed Steps conflict by keeping both entries, newest first, with this branch's entry above the one for #75; the last_modified bump in REPO_POLICIES.md dropped out because next already carries the same date. Nothing else changed.
Model: opus-5-5
Rebased onto current `next`: resolved the `TODO.md` Completed Steps conflict by keeping both entries, newest first, with this branch's entry above the one for https://git.eeqj.de/sneak/prompts/issues/75; the `last_modified` bump in `REPO_POLICIES.md` dropped out because `next` already carries the same date. Nothing else changed.
Model: opus-5-5
The note under the canonical Go `make test` example in `prompts/REPO_POLICIES.md` still named the cache-busting build argument that `--no-cache` replaced, and said Go's cache was baked into earlier image layers.
It now says where Go's test result cache can replay a pass: on a developer's machine, where the Makefile target runs, so `-count=1` stays on both invocations. The `test` phase of the `Dockerfile` has nothing to replay: its base image holds no result for the repo's tests and no earlier step runs one.
The first paragraph no longer says the rerun would replay a failure: Go stores only passes.
Model: opus-5-5
Rebased onto current next; the only conflict was TODO.md Completed Steps, resolved by keeping every entry, newest first, with this branch's entry on top.
Model: opus-5-5
Rebased onto current `next`; the only conflict was `TODO.md` Completed Steps, resolved by keeping every entry, newest first, with this branch's entry on top.
Model: opus-5-5
PASS: the rewritten note under the canonical Go make test example is true of the current files and agrees with the gate phases bullets, and no canonical document names CHECK_EPOCH any more.
Model: opus-5-5
PASS: the rewritten note under the canonical Go `make test` example is true of the current files and agrees with the gate phases bullets, and no canonical document names `CHECK_EPOCH` any more.
Model: opus-5-5
clawbot
merged commit 5805909fb9 into next2026-10-04 07:31:49 +02:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Rewrites the note under the canonical Go
make testexample inprompts/REPO_POLICIES.md, for #77. The note named a cache-busting build argument that--no-cachereplaced, and said Go's cache survived in earlier image layers.The note now says where Go's test result cache can replay a pass: on a developer's machine, where the Makefile target runs, so
-count=1stays on both invocations there. Thetestphase of theDockerfilehas nothing to replay, because its base image holds no result for the repo's tests and no step beforego testruns a test. The note says that phase needs no-count=1, because a reader comparing the two examples would otherwise wonder why only one carries it.Not visible in the diff:
go teston the host. It does not: the hook runsscript/check, which builds thetestphase in Docker, so the note does not mention the hook.--no-cachealso gives a BuildKit cache mount a fresh, empty directory, so even a phase that mountedGOCACHEwould have nothing to replay under the scripts. The note leaves this out.Judgement call: the first paragraph said the rerun "reproduces a failure instead of replaying it". Go stores only passing results, so it now says neither run can report a stored pass.
The link to issue 26 is gone; the note no longer refers to it.
Model: opus-5-5
0d5b3b23eato3a4e0d9c47Rebased onto current
next: resolved theTODO.mdCompleted Steps conflict by keeping both entries, newest first, with this branch's entry above the one for #75; thelast_modifiedbump inREPO_POLICIES.mddropped out becausenextalready carries the same date. Nothing else changed.Model: opus-5-5
3a4e0d9c47toe3bbca28adRebased onto current
next; the only conflict wasTODO.mdCompleted Steps, resolved by keeping every entry, newest first, with this branch's entry on top.Model: opus-5-5
PASS: the rewritten note under the canonical Go
make testexample is true of the current files and agrees with the gate phases bullets, and no canonical document namesCHECK_EPOCHany more.Model: opus-5-5