Devlog: H6, R4, H15 and H16

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-11-14 15:28:21 -05:00
co-authored by Claude Opus 5.5
parent 6873ba4ab4
commit e575bf0d42
+88
View File
@@ -1377,3 +1377,91 @@ no-op and the tree was clean. Nothing in the Agents category was addressed to th
issue's checklist (menus, drop, biomes, building, combat, tide phases, results), file
defects as `type/bug` and fix what's in reach. Look for placeholder art and text:
ARCHITECTURE.md still calls the figures "stand-in shapes".
## 2026-11-14 (afternoon, fourth run): H6 and R4 done, two visual bugs fixed
**Start of run (14:03 ET Saturday):** took the lock fresh. The pull was a no-op and the
tree was clean. The Agents category had a new topic (832) describing the agent mesh
across Eric's machines. It's informational and nothing in it is addressed to this task.
Eric was running the game on his own display during the run. That process wasn't
started by this task and wasn't touched.
**H6 (#59), closed: full visual QA pass.**
- Rendered all 82 capture scenarios and the HUD at 3840x2160, and looked at every one.
- `docs/VISUAL_QA.md` holds the checklist, with this pass's verdicts. It covers:
- menus, the lobby and the drop
- every biome and point of interest
- building, players and items, combat, the HUD and the chart
- each tide phase, the gale and under water
- results
- `make qa-captures` (`tools/qa_captures.sh`) renders the whole set one container at a
time. It waits whenever available memory drops under 1.5 GB (it was near 1 GB at the
start of the run). `tools/contact_sheet.gd` lays the captures out four to a sheet.
The full pass took about 40 minutes.
- No placeholders in what ships. ARCHITECTURE.md's stale "stand-in shapes" line is
gone. The plain shapes in the `camera*` scenarios are test cards under `test/`, which
isn't exported.
**H15 (#84), filed and fixed: the late match drawn in the underwater murk.**
`MatchView.setup()` syncs with the camera at the origin, which is under the sea once it
has risen. Nothing re-checked the murk when the camera was then placed, so:
- `match_tide` was a flat teal frame;
- `match_final` and `match_results` drew dry land in murk;
- in the game, every camera cut showed one frame of the wrong murk, because
`ClientSession` syncs before it moves the camera.
`_focus()`, which every camera placement calls, now applies the weather again.
**H16 (#85), filed and fixed: the world's square edge showing in the sea.**
- **Symptom:** the lobby view showed a pale square round the island, the terrain grid's
edge.
- **Cause:** the water shader's depth came from the view-space ray, so the open floor
(12.4 m down) read as only 7 to 11 m from the air. On top of that, the depth curve
never quite reached "deep", and the floor showed through.
- **Fix:**
- The shader takes the larger of the true vertical depth (from the world-space ray)
and the old estimate. Shallows at grazing angles keep the old estimate.
- Water is fully `SEA_DEEP` by `WaterView.OPEN_DEPTH` (9 m).
- Fully deep water is opaque.
- **Checked:** the square is gone in both Forward+ and Compatibility. The shallows still
read as shallow, and the open sea is a truer deep blue.
**R4 (#63), closed: originality and licence audit.**
- `docs/ASSET_PROVENANCE.md` traces all 162 files under `assets/`. It's written by
`tools/provenance.gd`, and the new `test_asset_provenance` keeps it complete and
current. The tarball and the .deb ship it.
- `git grep` found no commercial game names, characters, items or places. The only hits
were the genre wording, plus one open-source engine mentioned here in the DEVLOG.
- The engine's licence and its third-party notices already shipped in both packages.
There are no other dependencies.
- Name check: no "Tidewall" game or mark turned up. The near neighbours are Tidewell (a
seaside builder), Tidewild and Tideward (a board game). Noted on R7 (#66) with a
recommendation to run a USPTO search. Not legal clearance.
**Tests:**
- `make check` passed (696 tests).
- `make dist-check` and `make deb-check` passed after the packaging change.
- Every capture named above was re-rendered after its fix and viewed.
**Decision:** R5 is unblocked: R1 to R4 and H2 to H5 are all closed. I left it for a
fresh run, because it needs `make check`, `make match` and a full 30-minute `make soak`
on the tagged commit, then the release artifacts downloaded and verified in a clean
container. That doesn't fit the end of this run.
**Pace (15:30 ET Saturday):**
- About 40 hours remain to Monday 08:00.
- Release path left: R5. R6 and R7 wait on Eric. #72 (launch fiction) is also his.
- Still open and not blocking: H13 (an unreproduced soak hash mismatch; R5's soak will
show whether it recurs), and the deferred low-priority items (H8, H10, H14, B4, A3)
and features (M6, M7, A12, A13).
**Next run:** R5.
1. Run `make check`, `make match` and `make soak` on `main`.
2. If all pass, tag `v1.0.0`, build `make dist` and `make deb`, and create the Gitea
release with the tarball, the .deb and SHA-256 checksums. The server is the same
binary and is in the tarball.
3. Download the artifacts back, verify the checksums, and run them in a clean
`tidewall-` container.
4. Write release notes saying what isn't done and that nothing was tested on real GPU
hardware.
5. If the soak fails (H13), don't cut the release: report it.