Skip to content

release/0.4: backport #81, restoring the retractions v0.4.2 lost - #91

Merged
antoinetoussaint-byte merged 1 commit into
release/0.4from
backport/81-retractions-0.4
Oct 26, 2026
Merged

antoinetoussaint-byte merged 1 commit into
release/0.4from
backport/81-retractions-0.4

Conversation

@antoinetoussaint-byte

Copy link
Copy Markdown
Contributor

Summary

Backports #81 (e73dfc6) to release/0.4 with cherry-pick -x.

Why: v0.4.2 lost #81's retractions. #81 added two retractions by name for pseudo-versions that carry the deleted second Work Context implementation and sort outside [v0.1.51, v0.1.65]: v0.1.66-0.20260917202218-97dfea9e20d8 and v0.1.51-0.20260927165945-8157d3abecb2. Released modules pinned them from the former compat/v0.1.65 branch. v0.4.1, cut on main, carries those retractions. release/0.4 branched at v0.4.0, before #81, so v0.4.2 (cut on release/0.4) does not. Go reads retractions from the go.mod of the highest version, which is v0.4.2, and that go.mod retracts only [v0.1.51, v0.1.65]. The two pseudo-version retractions v0.4.1 published are therefore not in effect today.

The same gap is why release/0.4's CI is red. Its TestEveryPublishedTagCarryingTheImplementationIsRetracted reads every git tag, and the reachability tag compat-v0.1.65-pins is not a version ("compat-v0.1.65-pins" carries a prerelease and this comparison cannot order one). #81 is the real fix: the gate reads only vX.Y.Z tags (semver.IsValid), and go.mod retracts what that tag keeps reachable. Changing the test alone would make the gate green while the retractions stay lost.

Conflict resolution: only the go.mod require line and the matching go.sum lines conflicted. I kept release/0.4's google.golang.org/grpc v1.84.0 (and its golang.org/x/exp sums) and added golang.org/x/mod v0.42.0. go mod tidy is clean afterwards. The retract block and published_tags_test.go applied without conflicts.

Next: once this merges, #89 (#87 + #88) and the backport of #90 land on release/0.4, and then v0.4.3 is cut there so its go.mod, the new highest, publishes the full retraction set again.

Test plan

On this branch:

  • go test -race -count=1 ./... (root) passes, including TestEveryPublishedTagCarryingTheImplementationIsRetracted, which fails on the release/0.4 tip df5bd6a.
  • cd workcontext && go test -race -count=1 ./... passes.
  • Coverage: root 79.4% (gate 75).
  • ./scripts/check-one-implementation.sh: ok. ./scripts/security-posture.sh: SP-WC-01, SP-WC-05 ok. golangci-lint: 0 issues.
  • go mod tidy leaves go.mod and go.sum unchanged.

Fix or hack

Fix: a reviewed change from main, backported with only the dependency line reconciled.

🤖 Generated with Claude Code

…rsions released modules pinned from compat/v0.1.65 are retracted by name (#81)

The compat/v0.1.65 branch was replaced by the tag compat-v0.1.65-pins (annotations 0.2.15, robin 0.1.23 and record-store 0.2.2 pin two of its commits and must keep building). Go reads only vX.Y.Z tags as versions, so the published-tags gate now reads only those; what that tag keeps reachable is two pseudo-versions carrying the old Work Context implementation, outside the [v0.1.51, v0.1.65] retraction — retracted here by name.

Co-authored-by: Antoine Toussaint <antoine.toussaint.subscription3@obin.ai>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit e73dfc6)
antoinetoussaint-byte added a commit that referenced this pull request Oct 26, 2026
…on the restored retractions

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@antoinetoussaint-byte
antoinetoussaint-byte merged commit 71fd0f4 into release/0.4 Oct 26, 2026
16 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant