Nozzle and rain count only their own grains in flight; factory launches uncapped (t95)

This commit is contained in:
2026-10-29 18:12:47 -04:00
parent 3b427aeecc
commit d936f65553
5 changed files with 62 additions and 3 deletions
+6
View File
@@ -610,3 +610,9 @@ above, including:
cells. Exclude dedicated construction materials from give, discovery and blowing. On load,
clear loose construction grains and their bag/nozzle inventory; preserve owned blocks,
slope seals and machines. Ordinary wall ingredients still become rubble, scrap or bars.
- **Separate airborne allowances (t95, 2026-10-28).** The owner accepts a nozzle cap but
factory production must never stall because of a shared flyer limit. Only bag-blown grains
count toward the raised nozzle allowance (4,000); only rain counts toward rain's 2,500.
Factory launches remain uncapped. A full nozzle allowance says “Too much in the air.”
The 4,000 tuning preserves the existing eight-second blown-pile collapse check; 6,000 did not.
+14
View File
@@ -4,6 +4,20 @@ What agents have done, newest first. Add an entry for every piece of work: the d
note or request it was for, what changed, the files touched, the tests run, and anything left
open or not verified. Keep each entry short; the git history has the details.
## 2026-10-29: t95 separate nozzle and rain flight allowances (Codex)
- Nozzle counts bag flyers only, up to 4,000, with “Too much in the air.” feedback;
rain counts only rain, up to 2,500. Factory launches remain uncapped. No worldgen/save changes.
- Files: src/sim/{vacuum,weather}.js, test/t_nozzle.js, docs/DESIGN.md, tasks/t95* and this log.
- Eight new checks failed first; all 37 headless suites passed (19 nozzle checks).
Initial 6,000 allowance failed existing collapse checks; 4,000 passes without physics/test changes.
Only browser test: /tmp/playwright-test-t95.py (held right click, limit toast/resume, busy factory,
no page errors); inspected shots/t95-nozzle-limit.png (pixel). Sandbox startup stalled;
outside-sandbox GPU run passed after correcting test setup. No full browser suite.
- CPU flyTick median/p95 at 20,000 flyers: before 0.40/0.45 ms, after 0.25/0.33 ms;
timings are noisy and update code is unchanged. Combined 100 blow/rain calls: 7.6 ms after.
Logs: /tmp/t95-{headless-final,browser-final,perf-final}.log. Owner playtest pending. No commit.
## 2026-10-29: t97 keep construction grains out of the bag (Codex)
- PARTMAT now blocks give/discovery, inventory/refunds and single/mixed blowing. Load clears