Repository navigation
ci: run the scaffold tests on the full CI, and on a candidate pull request only when it touches packages/create - #151
Merged
Conversation
…the scaffold's .git
…, and what --list shows
ilmeskio
marked this pull request as ready for review
October 21, 2026 11:30
Merged
This was referenced Oct 21, 2026
ilmeskio
added a commit
that referenced
this pull request
Oct 21, 2026
…quest only when it touches packages/create (#151)
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
The scaffold level of the tests (
packages/create: a project generated and committed with git,@lntt/createpacked with npm, and the smoke that installs, lints and typechecks a generated project) leaves the minimal CI, unless the change touchespackages/create:ci.yml: one test step, its condition written once. The full CI (base the default branch, or a push) runspnpm testwithSCAFFOLD_SMOKE=1, as before. A pull request into a candidate runspnpm testwhengit diff origin/<base>...HEAD -- packages/createis not empty,pnpm test:minimalotherwise.pnpm test:minimal(rootpackage.json): every test but@lntt/create's.scaffold.test.ts: the residue check (grep -rl) leaves out.git. A git process the scaffold leaves behind could still rewrite.git/objectswhile grep read it, and the test failed only underci:local's load: reported in feat: queue ci:local runs on a lock and time out each step with the machine's load #150 asgrep -rl __LNTT_finding.git/objects/*entries vanishing while it reads, 1 run in 3 oncandidate/5's ownlntt ci local.docs/process/ci.mdstates the rule and its reason, with the scaffold level on every candidate pull request among the discarded directions; theci.ymlheader and the README say the same.Why: the scaffold test is a smoke test, slow and machine-dependent; its red blocked every child of the lntt-v0-6-2 theme at once for reasons unrelated to their changes (the person's decision). A scaffold failure a pull request does not see is caught on the candidate and fixed forward there.
Definition of done
pnpm ci:local --base candidate/5 --list: the test step and its condition; on fd5e8e3 (nopackages/createchange in the comparison at the time) the minimal CI ranpnpm test:minimalin 40.4 s with nopackages/createoutput; on c936402 (this branch touchespackages/create) it ran@lntt/create's 9 tests, the smoke skippedpnpm ci:local --base candidate/5on b4d29d4: 1 passed, 0 failedscaffold.test.tsalone: 3 runs in 3 green with the changelntt review: 2 rounds (541k, 643k tokens). Round 1's blocker (the record said every test runs on every run) fixed by rewriting the rule indocs/process/ci.md; its warnings fixed (one step, one condition; headers and docs aligned). Round 2: f14 answered and withdrawn (the smoke was already full-CI only), f8 answered and withdrawn, notes fixed. f7 kept by its reviewer: no test fails iftest:minimalstops excluding@lntt/create; the answer (a test of a script's text restates it, and the effect is only a slower minimal CI visible in the log) was not accepted. The person accepted the note as it is (2026-10-21): the risk is a slower minimal CI, never a defect that passes.🤖 Generated with Claude Code