Uh oh!
There was an error while loading. Please reload this page.
chore(release): bump VERSION to 0.10.13 for next cycle - #569
Conversation
Declares `0.10.13` as the next release version so the release train can cut an rc for it. `v0.10.12` is already tagged -- both `v0.10.12` and `v0.10.12-rc.1` exist -- and the delta waiting on `develop` touches `internal/doctor/doctor.go`, under the `internal/*` publish_path in release-train's `repos.yml`: - #563 fix(release): retry the cosign call, bounded and transient-only - #564 feat(doctor): measure disk, the dimension it never looked at With `VERSION` still reading `0.10.12`, the staging hop's version preflight refuses: rc NOT tagged - v0.10.12 already exists and the delta DOES touch published files (bump VERSION on develop before the next release) Measured on the 2026-08-25 staging hop: the `cli` leg failed this preflight, so `cli` sat the hop out. Same shape as the 0.10.12 bump (#562) one day earlier -- each hop that ships published files needs its own bump, and the previous one is consumed by the hop that shipped it. Patch-level, matching the whole 0.10.x series. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
saqlainsyed007
left a comment
There was a problem hiding this comment.
Approving. I checked the standards concern (org rule: "never hand-bump a version file") against release-train's own code, and this is the sanctioned mechanism rather than a bypass: scripts/promote-repo.shreadsVERSION and refuses to cut an rc when v$ver is already tagged ("NOT tagged - v$ver already exists" — the exact message, pinned in promote-verify.sh). It does not bump VERSION itself, so declaring the next cycle's version on develop is the input the train needs. develop isn't a publish branch, no tag is cut here, and the train still owns the actual rc/promotion — so this doesn't cross the hand-cut-tag / hand-publish line the rule is about.
Correct and unblocks #563/#564's internal/doctor delta for the next rc. CI green.
Uh oh!
There was an error while loading. Please reload this page.
Summary
Declares
0.10.13as the next release version so the release train can cut an rc for it.v0.10.12is already tagged — bothv0.10.12andv0.10.12-rc.1exist — and the delta waiting ondeveloptouchesinternal/doctor/doctor.go, under theinternal/*publish_path in release-train'srepos.yml:fix(release): retry the cosign call, bounded and transient-onlyfeat(doctor): measure disk, the dimension it never looked atWith
VERSIONstill reading0.10.12, the staging hop's version preflight refuses:Measured, not inferred: the
clileg hit this on the 2026-08-25 staging hop andclisat the hop out while the rest of the fleet rode. It cannot ride any hop until this lands.Same shape as #562 (the
0.10.12bump) one day earlier. Worth stating the pattern out loud: each hop that ships published files needs its own bump, because the previous bump is consumed by the hop that shipped it. This is the second consecutive day theclileg has been the only repo held for this reason.Patch-level, matching the whole
0.10.xseries.Type
chore(release)
Test plan
No behaviour change.
VERSIONis read byrelease.ymlat tag time and by the release train's version preflight; nothing in the binary reads it at runtime.Verified nothing else in the repo pins the old version:
File shape is byte-exact — single line, trailing newline preserved (
0.10.13\n), confirmed withxxd— which is whatrelease.ymlexpects.Checklist
developVERSION, 1 line)Note
Low Risk
Single-line version stamp only; no application logic, security, or data-path changes.
Overview
Bumps the repo
VERSIONfile from0.10.12to0.10.13so the release train can tag the next rc afterv0.10.12is already cut.This is a release-metadata-only change: it unblocks staging/version preflight when
develophas deltas under published paths (e.g. doctor and release fixes queued after0.10.12), without altering runtime CLI behavior.Reviewed by Cursor Bugbot for commit 26fa42c. Bugbot is set up for automated code reviews on this repo. Configure here.