You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
The version declarations are one merge behind the v0.3.44 tag #2768
The release chain reads the version from the tree, so the next tag cannot be cut from a tree that claims to be the previous release: the declarations must move to 0.3.45 first, and the tag follows only once that PR has landed and master's Test on the merge commit has concluded green.
What this asks for
Every version declaration moved to 0.3.45 in one change — bump-version.py rewrites all eight, and tests/test_version_sync.py proves they agree.
Nothing else in the tree touched: the diff is the eight declarations and no behaviour.
Not merged in the cycle that opens it, and no tag in that cycle either: the tag is cut from the merged tree, not from the branch.
Reopened: the merge that landed this issue's version bump is dedadbba, and its tree is byte-identical to the tree the bump's suite was measured on, so the declarations do read 0.3.45 on master. What the issue is actually about is still ahead of it: nothing is tagged for 0.3.45 yet, and no release exists.
Still owed, in order:
a Test run on the exact merge commit dedadbba read to a green conclusion on both legs;
an annotated tag at that commit, pushed;
Build Release read green per leg, including macOS signing and notarization, which no other reading covers;
check-release-published.py answering that the release is published, not a draft, not a prerelease, with the full asset set, and read as Latest.
This closes on that last reading, not on the merge.
v0.3.45 is published, so what this issue still owed is now measured rather than promised.
Tagv0.3.45 is an annotated tag at dedadbbaea38911b18ffd3c39f37297f13dcb7ca — the merge of the 0.3.45 bump. Remote reads tag object 43aa783b, dereferencing to that commit.
The merge landed a tree its suite had measured. Master's tree after the merge is byte-identical to the landing tree the suite ran on, so the 6186 passed / 37 skipped reading applies to master exactly rather than to a neighbouring tree.
Both host-side preflights answered before the tag was pushed:bump-version.py --check — all 8 declarations agree on 0.3.45; check-release-tag.py v0.3.45 — the tag names the version the tree declares.
The Test run on the exact merge commit was read green on both legs (test, test-windows), on run 42585058116. The first run on that same commit (42584346477) was red on one leg and the red was not about the tree: the actionlint tarball download answered HTTP 504 three times, which the step reports as an upstream fault instead of skipping. The re-triggered run on the same commit is what the tag rests on — the commit never moved, so "exactly this commit" still holds.
Build Release run 42585826268 is green on every leg, with elapsed times: ubuntu x86_64 08:25:42→08:35:17, macos-15 arm64 08:25:45→08:35:28, ubuntu aarch64 08:25:43→08:36:01, windows-2025 x64 08:25:42→08:37:26, and the release job 08:37:28→08:38:11. macOS signing and notarization are covered by that leg and nowhere else.
scripts/check-release-published.py v0.3.45 exits 0: not a draft, not a prerelease, 11 assets, and GitHub reads v0.3.45 as Latest (v0.3.44 now reads latest=false). The asset set matches v0.3.44's shape name for name: both AppImages with their zsync files, both linux tarballs and .run installers, the macOS pkg, the Windows exe and zip.
The merge alone did not finish this issue — the bump's Closes closed it one second after landing, while the tag, the build and the published release were all still ahead of it. That is why it was reopened and is closed here instead, on the outward reading.
Activity
argszero commentedon Nov 9, 2026
Handled by #2769
how2how2how2-arch commentedon Nov 9, 2026
Reopened: the merge that landed this issue's version bump is
dedadbba, and its tree is byte-identical to the tree the bump's suite was measured on, so the declarations do read 0.3.45 on master. What the issue is actually about is still ahead of it: nothing is tagged for 0.3.45 yet, and no release exists.Still owed, in order:
dedadbbaread to a green conclusion on both legs;check-release-published.pyanswering that the release is published, not a draft, not a prerelease, with the full asset set, and read as Latest.This closes on that last reading, not on the merge.
how2how2how2-arch commentedon Nov 9, 2026
v0.3.45 is published, so what this issue still owed is now measured rather than promised.
v0.3.45is an annotated tag atdedadbbaea38911b18ffd3c39f37297f13dcb7ca— the merge of the 0.3.45 bump. Remote reads tag object43aa783b, dereferencing to that commit.bump-version.py --check— all 8 declarations agree on 0.3.45;check-release-tag.py v0.3.45— the tag names the version the tree declares.test,test-windows), on run 42585058116. The first run on that same commit (42584346477) was red on one leg and the red was not about the tree: the actionlint tarball download answered HTTP 504 three times, which the step reports as an upstream fault instead of skipping. The re-triggered run on the same commit is what the tag rests on — the commit never moved, so "exactly this commit" still holds.scripts/check-release-published.py v0.3.45exits 0: not a draft, not a prerelease, 11 assets, and GitHub reads v0.3.45 as Latest (v0.3.44 now readslatest=false). The asset set matches v0.3.44's shape name for name: both AppImages with their zsync files, both linux tarballs and .run installers, the macOS pkg, the Windows exe and zip.The merge alone did not finish this issue — the bump's
Closesclosed it one second after landing, while the tag, the build and the published release were all still ahead of it. That is why it was reopened and is closed here instead, on the outward reading.