feat(ci): turn pre-publish layer 2 on by default - #36
Merged
Conversation
`run-prepublish-install` now defaults to `true`. The mechanism shipped finished and measured in #35; what remained was a policy call with a real blast radius, and it was made on 2026-08-05. As of that date no caller passes `run-prepublish-install`, and none passes `expect-unpublished-deps` either, so the default is the whole story: all thirteen callers pick up layer 2 and the default allowance with no change on their side. Measured the same day against the live registry with that allowance: 11 pass, 2 blocked-peer (`transform`, `synth`, on `@cosyte/fhir`), 0 red. Both blocked-peer results exit 0, so the flip reds nobody. It is recorded as a dated measurement rather than a standing guarantee, in the workflow comment and in the README both, because the registry is not ours and an answer that holds today can change under us. `blocked-peer` still establishes less than "fully explained": npm short-circuits, so a second defect hiding behind a declared absence stays invisible. That gap is pre-existing, already set out under "The verdicts", and this change neither widens nor narrows it. The README follows the code in the same commit. The layer table reads `on`, the adoption note becomes an opt-OUT example, and the paragraph that told a caller to add `run-prepublish-install: true` is gone, because after the flip there is nothing to adopt. The test that pins the two defaults is flipped with them and re-anchored at the six-space declaration. The new comment block spells `run-prepublish-install: false` while explaining how to opt out, and an unanchored search would match that decoy first and then scan forward into the real declaration, reading the right value from the wrong place.
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
1 participant
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.
PREPUBLISH-INSTALL-LAYERS, final step.run-prepublish-installnow defaults totrue.Both layers shipped finished and measured in #35. Layer 2 shipped default OFF on purpose: turning it
on for thirteen repos at once is a policy call with a real blast radius, not a build. That call was
made on 2026-08-05. This PR is the flip and nothing else.
Blast radius
As of the flip no caller passes
run-prepublish-install, and none passesexpect-unpublished-depseither, so the default is the whole story: all thirteen callers pick up layer 2 and the default
allowance with no change on their side. There is nothing for a caller to adopt.
Measured on 2026-08-05 against the live registry with the default allowance: 11
pass, 2blocked-peer(transform,synth, on@cosyte/fhir), 0 red. Bothblocked-peerresults exit0, so the flip reds nobody. Re-measured independently during review and it reproduced: 11 install
clean, exactly 2 fail (
transformE404,synthERESOLVE), both on@cosyte/fhir, which thedefault allowance declares
=blocked.That is recorded in the workflow comment and in the README as a dated measurement, not a standing
guarantee. The registry is not ours.
What this does not claim
blocked-peerestablishes less than "fully explained". npm short-circuits, so a second defecthiding behind a declared absence is invisible. That gap is pre-existing, already set out under "The
verdicts", and this change neither widens nor narrows it. The README wording was tightened so the
adoption note cannot be read as the stronger claim.
=blockedand=privatestay in different words, untouched:=blockedexcuses,=private(
assets,docs) excuses nothing in either gate.Docs follow code, in the same PR
The layer table reads
on, the adoption note becomes an opt-out example, and the paragraphtelling a caller to add
run-prepublish-install: trueis gone.Test
The test that pins the two defaults is flipped with them and re-anchored at the six-space
declaration, because the new comment block spells
run-prepublish-install: falsewhile explaininghow to opt out. Unanchored, a search matches that decoy first and then scans forward into the real
declaration, reading the right value from the wrong place. Mutation-checked during review: flipping
either default back, or deleting the declaration while leaving the decoy, all fail the test.
Gate:
actionlintexit 0,node --test "test/*.test.mjs"285/285.