Devlog: R5 gates, H17, v1.0.0 tag and draft release

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-11-14 18:36:51 -05:00
co-authored by Claude Opus 5.5
parent f5227b731d
commit 358234d246
+65
View File
@@ -1465,3 +1465,68 @@ container. That doesn't fit the end of this run.
4. Write release notes saying what isn't done and that nothing was tested on real GPU 4. Write release notes saying what isn't done and that nothing was tested on real GPU
hardware. hardware.
5. If the soak fails (H13), don't cut the release: report it. 5. If the soak fails (H13), don't cut the release: report it.
## 2026-11-14 (evening): R5 release gates, v1.0.0 tagged, release drafted
**Version 1.0.0 (f776d59).** `application/config/version`, the install commands in
PLAY.md and SERVER.md, and the README status now say 1.0.0. Also in the README:
- The effects limitation is gone, since A7 shipped.
- The Later issues for keg-throwing bots (#83) and list servers in the Join menu (#82)
are linked.
**First gate run on f776d59:**
- `make check` passed (696 tests).
- `make match` passed.
- `make soak` failed. Every client process "logged its memory at 1 match starts". Hashes
(2,438 compared, 0 differing), tick timing and server memory all passed.
**H17 (#86), filed and fixed in f5227b7.**
- **Cause:** match 0 of the seed-7 soak was decided during the drop. The 24 clients
fought as they landed, with 23 eliminations in 39 s, so the match never reached
play. That's allowed by design. `test_match` already ends a match in the drop.
- **The gap:** `soak_clients.gd` sampled memory only on entering play, so that match
gave no sample.
- **Fix:** such a match is now sampled at its results. `samples_memory_in()` is
unit-tested, and the memory line says where the sample came from. Earlier soaks
passed because their first match reached play.
**Gates rerun from scratch on f5227b7, all passing:**
- `make check`: 698 tests.
- `make match`: the replay hash matched.
- `make soak`: 2,458 hashes compared, 0 differing; p99 tick 6.7 ms; client growth
0.65 MB, server 0.39 MB.
- `make dist-check` and `make deb-check` passed. The packages' `VERSION` says
f5227b731d2e.
**Tag `v1.0.0`** (annotated, on f5227b7) is pushed.
**The Gitea release is a draft:**
- It has the release notes and both `.sha256` files.
- The Gitea tools only take uploads as base64 inside the call, which doesn't work for
a 30 MB tarball and a 22 MB .deb, and I didn't take a token from any config to
upload another way.
- The binaries are kept in the git-ignored `build/release-v1.0.0/`, which `make dist`
doesn't touch.
- R5 (#64) is labelled `needs-eric`. Eric attaches the two files and publishes (now or
after his playtest). Then a run downloads them anonymously, checks the checksums and
runs them in a clean container before closing R5.
**H13 (#80):** two more soaks with no mismatch, about 9,000 comparisons since its
diagnostics went in.
**Pace (18:40 ET Saturday):**
- The release is gated, tagged and drafted, 37 hours before Monday 08:00.
- Waiting on Eric:
- R5: attaching the binaries.
- R6: the playtest.
- R7: the name and the public server.
- #72: the launch fiction.
**Next run:**
1. If the release has its binaries, download them from the release, verify them and
run them in a clean `tidewall-` container. If all is well, close R5 and the v1.0
milestone.
2. Otherwise, take the deferred work, bugs first: H13 (#80), then H14 (#81). After that,
M6 and M7 (bots that build, duos), which carry no priority label.
- A change to the game after v1.0.0 would need its own point release. Keep `main`
releasable.