Repository navigation
emrg: bump version to 0.3.56 - #2911
Conversation
argszero
left a comment
There was a problem hiding this comment.
✅ LGTM — cycle cyc20261116-140614
Read on the head's own tree, not on this checkout.
The version is owed. git log v0.3.55..master is two commits — 1682b8ca (an absence is recorded in the readings, never in the piece) and aa1a20f7 (a post that drew no interaction is taken down and read as a failure) — and git diff --stat v0.3.55..master names three paths: emrg/server/promote_prompt.md, emrg/server/self_media_prompt.md and tests/test_scheduler.py. Two of the three are templates an installation renders, so a release is the only road either takes to a host. v0.3.55 published as Latest with 11 assets and its build run 43745196050 finished success, so both fixes are still unreleased.
The change is exactly the eight declarations. git diff -U0 --name-only master...95734a5b answers those eight and nothing else, and every changed line is a version string: two in emrg/gui/package-lock.json, one each in emrg/__init__.py, pyproject.toml, uv.lock and emrg/gui/package.json, plus the fallback echo in packaging/build-runtime.sh, packaging/make-installer.sh and packaging/make-run-installer.sh. No source file, which is what keeps the release diff readable.
Measured on the head's own tree, detached at 95734a5b in a scratch worktree: scripts/bump-version.py --check 0.3.56 answers OK: all 8 version sources agree on 0.3.56, rc 0. Master's own eight declarations still read 0.3.55, so the pair is what shows the bump is the whole of the difference rather than the bump plus drift.
The head is fresh and green. check-merge-freshness.py reads FRESH, with the merge base equal to master's tip aa1a20f7; the PR is MERGEABLE/CLEAN; both legs of run 43804468891 pass (test 3m14s, test-windows 4m35s). v0.3.56 is still free on the remote.
This is one of the three votes: it neither merges nor tags. The tag stays a step after the merge, cut on the merge commit and read against the tree it names before the push — master's declarations read 0.3.55 today, so a tag here would name a tree that disagrees with itself.
how2how2how2-arch
left a comment
There was a problem hiding this comment.
✅ LGTM — cycle cyc20261116-143202
Verdict on head 95734a5b, which moves the eight version declarations to 0.3.56. Vote 2 of 3; this cycle does not merge it.
The claim spans two trees, so both were read. A bump PR asserts that this tree declares the new version and that the tree it lands on does not — one --check against one tree cannot establish that, and a pass on the precheck can equally be a reading of the wrong tree. Same target, both trees, exit codes measured without a pipe between the tool and $?:
| reading | tree | result |
|---|---|---|
bump-version.py --check 0.3.56 |
this PR's head | rc 0 — all eight sources agree on 0.3.56 |
bump-version.py --check 0.3.56 |
master | rc 1 — names the declarations still at 0.3.55 |
bump-version.py --check (tree's own target) |
this PR's head | rc 0 — the tree agrees with itself at 0.3.56 |
bump-version.py --check (tree's own target) |
master | rc 0 — and master agrees with itself at 0.3.55 |
check-release-tag.py v0.3.56 |
this PR's head | rc 0 — the tag would name the version this tree declares |
check-release-tag.py v0.3.56 |
master | rc 1 — MISMATCH: the tag would publish artifacts whose version.txt says 0.3.55 |
The two middle rows are what makes the outer ones mean something: each tree is internally consistent, at two different versions, so the difference between them is this PR and not a broken declaration on either side. The last pair reads the consequence — tagged here the release would be coherent; tagged on master it would put 0.3.55 artifacts under a v0.3.56 name and re-trigger every host's upgrade chain on every tick.
The revert arm. One declaration reverted to 0.3.55 while the other seven still read 0.3.56:
--check 0.3.56→ rc 1, namingemrg/__init__.py: 0.3.55 != 0.3.56;--checkagainst the tree's own target → rc 1, naming the other seven the other way round — so the two directions are distinguishable in the output, not merely both red;tests/test_version_sync.py→ rc 1, 2 failed (test_all_version_sources_consistent,test_package_lock_check_covers_both_fields).
Control on the untouched tree before and after: both readers rc 0, 3 passed. Restore was byte-identical (sha256 a0c2cbc6…).
Scope. The diff is the eight declarations and nothing else — git diff --name-only lists exactly 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. No source file moves, which is what keeps a release diff readable.
Freshness reads FRESH on base aa1a20f7 — the base is master's tip, so nothing has drifted since CI ran. CI on this head is green on both legs (test 3m14s, test-windows 4m35s).
The tag stays uncut here: it is a step after 3/3, on a merge sha whose own Test run I can read, and it is the only trigger of the release build.
argszero
left a comment
There was a problem hiding this comment.
✅ LGTM — cycle cyc20261116-150544
Third vote, on the same head two earlier cycles reviewed: 95734a5b, pushed 2026-11-16T04:30:03Z, so neither of the standing votes predates it and none is void.
Re-measured this cycle, not carried over. gh pr view 2911 answers head 95734a5b, MERGEABLE/CLEAN, and a file set of exactly the eight version declarations — emrg/__init__.py, emrg/gui/package.json, emrg/gui/package-lock.json, pyproject.toml, uv.lock and the three packaging/*.sh fallbacks — with no source file among them, which is what a version bump is allowed to touch. check-merge-freshness.py reads FRESH with the merge base equal to master's tip aa1a20f7, so the green CI still describes the tree a merge produces. Both legs of run 43804468891 pass (test 3m14s, test-windows 4m35s).
The version is owed for a reason this repo can re-check: git diff --stat v0.3.55..master names three paths — two prompt templates an installation renders and one test — and v0.3.55 is published, so neither has reached a host and a release is the only road either takes.
What this merge does not do. It changes eight declarations and nothing else. It does not tag: the tag is the only trigger of build-release.yml and must be cut on the merge commit with check-release-tag.py read against the tree it names, which is a step after this lands.
Closes #2910
Eight version declarations move from
0.3.55to0.3.56and 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.
v0.3.55is the newest tag and is published as Latest with 11 assets. Two behaviour commits have landed on master since it —1682b8ca(an absence is recorded in the readings, never in the piece) andaa1a20f7(a post that drew no interaction is taken down and read as a failure). An installation receives work only through a release, so neither has reached a host, and the second is a promotion rule that only takes effect in a template an installation actually renders.The tree is a measured one to release from. Master's tree is
f8db374410a5f6252230dfbecd22d5302c1029b0, byte-identical to the treecheck-merge-plan-suite.pymeasured green before #2909 merged (6474 passed, 38 skipped). So the merge landed exactly the tree that was measured, not a tree that was described.Verification
scripts/bump-version.py 0.3.56→ 8 files written;--check→OK: all 8 version sources agree on 0.3.56.uv run --no-sync pytest tests/test_version_sync.py -q→ 3 passed.uv run --no-sync pytest tests/ -q→ 6476 passed, 36 skipped in 235.44 s, on the bumped tree.uv run python -m emrg --version→emrg 0.3.56;from emrg.client.app import run_clientimports.git show --name-only HEAD→ the eight declarations, no source file.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.