Repository navigation
Bump omarchy-mac's version only in a release's own pull request - #30
Conversation
maralcbr
left a comment
There was a problem hiding this comment.
Thanks @scottjones, I'm OK with this. #21, #32 and #33 all bumping 0.1.3 → 0.1.4 make the case on their own.
- I merged main into the branch to clear the conflict (
omarchy-mac/README.mdline 3, where main said0.1.3); I kept your sentence and main's other README edits. - Edge doesn't lose anything:
mac-releasere-pins the same version with the next pkgrel, which the omacom recipe already allows. Could the Releases section say when to cut a version bump and when to just re-pin with a pkgrel increment? Otherwise "a release" is undefined while fixes keep landing on edge. - Non-blocking: the gate accepts any value in a release PR, so a lower or non-pkgver version (
0.0.9,banana) passes. Nothing ships wrong, because the nextmac-releasepin refuses it. But by then it's on main, and only another release PR can fix it. Checking it here, with the same pkgver pattern andversion_newercomparison againstHEAD^1thattools/releaseuses, catches it in the PR that made the mistake. - Since contributors no longer pick the number, could feature PRs say whether they're patch, minor or major in the body? #12 went to 0.2.0 for a reason, and the releaser needs that.
- The Merge Danger list is out of date: #28 is merged, and #32 and #33 also bump. Those PRs only fail once their checks re-run after this lands.
- nit: AGENTS.md prefers a full if/else over an
exitin one branch, so theGITHUB_EVENT_NAMEearly exit could be the first arm of one if/elif/else.
maralcbr
left a comment
There was a problem hiding this comment.
@scottjones my approval still stands, and the last two weeks make the case better than I could: since Oct 12 main went from 0.1.3 to 0.2.3 through feature PRs, with #12, #32, #54 and #56 all bumping on Oct 24 and #32 bumping twice to stay ahead of the others.
To land it:
- Merge main in. The only conflict is
omarchy-mac/README.mdline 3, which now says0.2.3; your sentence wins. - Refresh Merge Danger: #12 and #28 are merged, and #21 and #33 are the open PRs that still bump and will fail the gate.
Still open from my Oct 12 review, none blocking:
- The gate accepts any value on a
release/omarchy-mac-<v>branch, so a downgrade or a non-pkgver string gets onto main and only another release PR fixes it. Checking it againstHEAD^1with the same pkgver pattern andversion_newerthattools/releaseuses would catch it in the PR that made it. - The Releases section should say when a release bumps the version and when
mac-releasejust re-pins with a pkgrel, and feature PRs should say patch/minor/major in the body so the releaser can choose. tests.ymlhas nomerge_grouptrigger and the script exits early on anything butpull_request; fine today, but worth knowing if we turn on a merge queue.
I'll merge once main is in and CI is green.
| changed=$(git diff --name-only HEAD^1 HEAD) | ||
| if ! grep -qx 'omarchy-mac/version' <<<"$changed"; then | ||
| echo "omarchy-mac/version is unchanged." | ||
| elif [[ $changed == "omarchy-mac/version" && ${GITHUB_HEAD_REF:-} == "release/omarchy-mac-$(<omarchy-mac/version)" ]]; then |
There was a problem hiding this comment.
This accepts any value from a release/omarchy-mac-<v> branch, downgrades included. A pkgver-pattern and version_newer check against HEAD^1 would stop a bad number before it reaches main. Non-blocking.
There was a problem hiding this comment.
Agreed. We'll check the new value against HEAD^1 with the same pkgver pattern and version_newer that tools/release uses, in a follow-up PR.
|
Thanks. Main is merged in (40824ea, and again at fe554ce for #58), with our README sentence kept where main had 0.2.3, and the Merge Danger now lists #21 and #33 as the open PRs that still bump. We'll take the release-branch check, the Releases wording on bump versus re-pin with feature PRs naming patch, minor or major in their body, and the merge queue note in a follow-up so this stays as you approved it. |
[Medium impact] The PR appears safe to merge. SummaryThis PR moves version changes into release-only pull requests.
scottjones explicitly deferred automatic bump-PR creation to a follow-up. scottjones also acknowledged that open pull requests containing version bumps must remove those bumps to pass the new gate. Reviews (1) · Last reviewed commit: "Merge main into policy/release-sets-vers..." · Reviewed by Greptile |
PRs that each bumped omarchy-mac/version kept colliding (#23 and #24 on 0.1.1, #21 and #28 on 0.1.3) or forgot to bump. Only a release reads the number, through the recipe's pkgver, so the release now sets it: in a pull request from release/omarchy-mac-<version> that changes only that file. A new gate fails any other pull request that touches the file. The README no longer repeats the number. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
fe554ce to
fcdc335
Compare
Summary
PRs that each bump
omarchy-mac/versionkeep colliding (#23 and #24 on 0.1.1, #21 and #28 on 0.1.3) or forget to bump. Only a release reads the number, through the recipe'spkgver, so the release sets it.The README stops repeating the number. A follow-up teaches
tools/release/mac-release --bump patch|minorto open the bump PR, with a manual Action for it.Evidence
After: the gate fails it, and passes the release bump PR, PRs that leave the version alone, and pushes to main. Checked against eight scratch cases on current main: a release branch whose name doesn't match the new version, or that changes another file too, fails; push and manual runs pass.
Merge Danger
Door: two-way
Open PRs that bump the version fail the gate until they drop the bump: #21, #33, #72, #90, #94, #97, #125, #131, #134, #135, #144, #145, #152, #155 and #157 today. #140 already leaves it alone and relies on this rule. PRs merged before this kept their bumps; main is at 0.2.6.
Blast Radius: process
🤖 Generated with Claude Code