Repository navigation
emrg: release v0.3.17 — eight commits since v0.3.16, including the GUI timestamp fix the host reported on 0.3.16 #2344
Description
Origin: host instruction — 2026-10-15T22:15:21+08:00, session s_260727_1103_866d: "not 3.14 — I mean: after all the rants are done, publish another new version" (translated from the Chinese original; resolved on this host with scripts/find-host-message.py, 2 hits — the daemon log and the session history — both at that instant).
That condition is now met: every emrg rant is completed — the last two closed 2026-10-17T18:50+08 after their PRs landed on master. This issue is the release those rants were waiting for.
Baseline, measured
- Latest release: v0.3.16, published 2026-10-16T14:51:21Z, not a prerelease.
masteris 8 commits ahead (gh api repos/argszero/emrg/compare/v0.3.16...master).- The host's install reads
~/.emrg/install/version.txt= 0.3.16, so none of the eight is in force on this machine yet.
| commit | change |
|---|---|
2742584d |
the API push fallback stops repeating work whose answer is already known (#2329) |
bf10e6e0 |
the README's /skills row stops splitting on a pipe inside a table code span (#2327) |
b766b666 |
the re-seal guard names every spelling of a hand-written codesign call (#2331) |
3e383fcb |
the watchdog test's clock is the script's, not the CI runner's (#2333) |
32fa3eef |
the abstain window gets its far bound, so a vote on a head pushed by a later cycle counts (#2335) |
bbbd75e9 |
a verdict is read from the row, not from the document its branch name is printed in (#2337) |
64ea85fa |
only a head the voting cycle itself pushed is abstained on (#2339) |
ccd950b4 |
the GUI rows carry the moment the daemon recorded, on live and replay alike (#2341) |
Why this release matters beyond the count
Two of the eight close host-reported defects:
ccd950b4is the GUI half of "every message needs a timestamp", first asked on 2026-10-09 and reported again on 0.3.16 by the host on 2026-10-17T09:56 — the TUI half had shipped, the GUI half never existed. It is merged and unreleased, so the host is still running the defect.64ea85fais the abstain clause the host ruled on 2026-10-16T22:18:15 (only the cycle that pushed a head may not vote). Its tools are on master; the delivered template in the install tree still prints the old, wider rule, so a cycle running here is told one rule while its own counter enforces another. Only a Release moves that copy.
Stages
- Bump —
scripts/bump-version.py 0.3.17rewrites every version declaration at once, andtests/test_version_sync.pyproves they agree. PR it, take three approvals from different cycles, squash merge. A release PR is never self-merged. - Tag —
git tag -a v0.3.17 && git push origin v0.3.17. The tag is the only trigger ofbuild-release.yml. - Verify —
scripts/check-release-published.py v0.3.17: green across the platform matrix, published as Latest (not draft, not prerelease) with the full asset set.
What this release does not do
- It does not carry emrg: the cycle reads extra_prompt, and §1.5 stops paying for a read it never finishes #2343, which is open and was rejected on its tree (a §1.7 sentence states a fixed defect as the current rule, and the new §1.5 step-0 command does not parse). A prompt change that is not merged cannot be released.
- It does not change any behaviour by itself: a version bump plus a tag.
Activity
argszero commentedon Oct 17, 2026
Handled by the bump PR — stage 1 of the three in the issue.
argszero commentedon Oct 17, 2026
Handled by #2345 — the bump PR (stage 1). The tag and the publish verification follow its merge.