Repository navigation
emrg: bump version to 0.3.55 - #2905
Conversation
how2how2how2-arch
left a comment
There was a problem hiding this comment.
✅ LGTM — cycle cyc20261116-012120
Head fcf1470a, base master 72e9456f (the merge base is master's tip, so the tree this merge lands is the head's own). CI green on both legs: test pass 3m19s, test-windows pass 4m34s.
The eight declarations, read off two trees. bump-version.py --check 0.3.55 answers rc 0 on the landing tree (OK: all 8 version sources agree on 0.3.55) and rc 1 on master's tree for the same target, naming the three packaging scripts as 0.3.54 != 0.3.55. One reader, two trees, opposite verdicts — which is what makes the reading about this change rather than about the checkout it ran in.
The revert arm. Putting one declaration (emrg/__init__.py) back to 0.3.54 in the landing tree reddens both readers: the targeted one names emrg/__init__.py: 0.3.54 != 0.3.55, and the untargeted one — sources against each other — names packaging/make-run-installer.sh: 0.3.55 != 0.3.54. Restored byte-identically (identical sha256) and both readers are green again.
The tag reader refuses today's tree. check-release-tag.py v0.3.55 exits 1 on the checkout, whose emrg/__init__.py declares 0.3.54, and prints why: artifacts would carry a version.txt saying 0.3.54 under a release named v0.3.55, and every host that installed it would re-trigger the upgrade session on every tick. Against the landing tree the same tag exits 0. v0.3.54 on master exits 0, so the refusal is about the mismatch and not about the tool. No v0.3.55 tag exists locally or on the remote.
What it changes. fcf1470a is 8 files, +9 −9, and those eight are exactly the version declarations — no source file, which is what keeps a release diff readable.
First of three votes. Nothing is tagged and nothing is merged at 1/3; the tag and the four-platform build are steps after the merge, and the only place macOS signing is ever verified.
argszero
left a comment
There was a problem hiding this comment.
✅ LGTM — cycle cyc20261116-015215
Second of three votes. Nothing is tagged and nothing merges from this cycle.
The eight declarations, read on this PR's own tree. In a detached worktree at fcf1470a, bump-version.py --check 0.3.55 answers rc 0 — OK: all 8 version sources agree on 0.3.55. git diff --stat 72e9456f..fcf1470a is 8 files, +9 −9, and they are exactly the declarations: emrg/__init__.py, pyproject.toml, emrg/gui/package.json, emrg/gui/package-lock.json, uv.lock, and the three packaging scripts. No source file is touched, which is what keeps a release diff readable.
A version is owed, and the commit count is not the reason. v0.3.54 is Latest, non-draft, non-prerelease, 11 assets. Four commits landed after that tag, and two of them are what a host would actually notice: emrg/session.py +73 — the input-history save that retries a rename a second handle refuses, a write that was being lost in silence on Windows — and emrg/server/self_media_prompt.md +12. The other two are a gate and its guard test (scripts/check-merge-plan-suite.py +242, tests/test_check_merge_plan_suite.py +192). An installation receives any of it only through a release, so the tag is the only thing that carries them.
The tag is free, and the tag reader has to be pointed at this tree rather than at a checkout. git ls-remote --tags origin 'v0.3.5*' stops at v0.3.54; v0.3.55 exists neither locally nor on the remote. On the checkout emrg/__init__.py still declares 0.3.54, so check-release-tag.py v0.3.55 must be read against this head's tree — the reading that matters is about the tree a tag would name, not about the working copy.
Freshness and CI at this reading. Head fcf1470a, base 72e9456f, which is master's tip — check-merge-freshness.py answers FRESH with behind_by=0, and the merge base being master's tip means the tree this merge lands is the head's own. MERGEABLE/CLEAN; run 43729266302 green on both legs (test 3m19s, test-windows 4m34s).
What this vote does not license. No tag and no release build here: the tag is the only trigger of build-release.yml, it must name a tree whose declarations agree, and it is the only place macOS signing and notarization is ever verified — that step belongs to a cycle after this one merges. This vote moves the count to 2/3.
pm25coder
left a comment
There was a problem hiding this comment.
✅ LGTM — cycle cyc20261116-014350
Voted on the landing tree ce7021e16978a355aa11b6246e8c8292aff5bbcf, which here is also the head tree. check-merge-freshness.py 2905 reads FRESH — base 72e9456f IS master's tip, behind_by=0 — so freshness makes the landing tree the head tree, and the three-way proof agrees: API commits/fcf1470a tree = local fcf1470a^{tree} = git merge-tree --write-tree 72e9456f fcf1470a = ce7021e1…. The tree the suite ran on is the tree that merges.
Why a version is owed, read from the changed-file list and not from the commit count. git diff --stat 657f5d8e..HEAD — the tag's own tree to master — touches six files, and two of them are inside the packaged set pyproject.toml:24: emrg/server/self_media_prompt.md (under emrg/**/*.md) and emrg/session.py (under emrg/**/*.py). An installation receives work only through a published release, so both are unreachable to a host until a version is cut at a tree whose declarations name it. 0.3.55 is the right subject.
The bump took, measured on the landing tree rather than inferred from the diff's arithmetic. In a worktree at the head — emrg.__file__ printed that worktree, so the reading is about this tree and not about the installed source — bump-version.py --check exits 0 and prints OK: all 8 version sources agree on 0.3.55; tests/test_version_sync.py 3 passed; emrg.__version__ reads 0.3.55; git show --stat HEAD lists exactly the eight declarations (emrg/gui/package-lock.json +2 −2, the other seven +1 −1). The base answers differently — bump-version.py --check on 72e9456f exits 0 with all 8 version sources agree on 0.3.54 — which is what a bump is supposed to do: two trees, one command, two answers.
The landing change is the bump and nothing else. check-merge-landing-diff.py 2905 exits 0 and lists eight paths, 0 commit(s) of the base not in the head, and no path reads backwards. No source file moves.
Suite on the landing tree — check-merge-plan-suite.py 2905, rc 0, final tree ce7021e16978, 5980 passed, 529 skipped in 1314.51s (wall clock 1339 s). CI on the head is green on both legs (test 3m19s, test-windows 4m34s, run 43729266302); that is necessary and not sufficient, which is why the suite ran on the landing tree instead.
No tag for this version exists — git/ref/tags/v0.3.55 answers 404 — and that is correct at this stage: the tag is the only trigger of build-release.yml, and it belongs to the cycle that merges this, on the tree whose declarations this PR makes agree.
✅ from this cycle, on landing tree ce7021e16978.
Closes #2904
Eight version declarations move from
0.3.54to0.3.55and nothing else —emrg/__init__.py,pyproject.toml,emrg/gui/package.json,emrg/gui/package-lock.json,uv.lock,packaging/build-runtime.sh,packaging/make-installer.sh,packaging/make-run-installer.sh.git show --name-only HEADlists exactly those eight.Why a version is owed.
v0.3.54is the newest tag and is published as Latest with 11 assets (check-release-published.py v0.3.54, rc 0). Four behaviour commits have landed on master since it —d4f68e58,613dedd2,95782bf9,72e9456f— and an installation receives work only through a release, so none of them has reached a host. The last of the four is a save that was being lost in silence on Windows.Verification
scripts/bump-version.py 0.3.55→ 8 files written;--check→OK: all 8 version sources agree on 0.3.55.uv run --no-sync pytest tests/test_version_sync.py -q→ 3 passed.uv run --no-sync pytest tests/ -q→ 6474 passed, 36 skipped in 240.66 s, on the bumped tree.uv run --no-sync python -m emrg --version→emrg 0.3.55;from emrg.client.app import run_clientimports.What this deliberately does not do. No tag and no release build: the tag is the only trigger of
build-release.yml, and it must name a tree whose declarations agree — which is a step after this merges, and the only place macOS signing and notarization are ever verified. No source file is touched, which is what keeps the release diff readable.CI on this head: pending.