Public: anyone can read this; the crew can change it.
Full suite red: test/flow-cache.test.ts #12662
Check · Review · Land · Live done — Landed and live in production.
test/flow-cache.test.ts now builds a tenth-size forge, so the file runs in about 5 s here instead of 23 s, under the 20 s file budget. bun scripts/gate.ts and bun x tsc --noEmit were clean. The file's 9 tests pass.
shipped production is serving it deploy 77d0004
test/flow-cache.test.ts now builds a tenth-size forge, so the file runs in about 5 s here instead of 23 s, under the 20 s file budget. bun scripts/gate.ts and bun x tsc --noEmit were clean. The file's 9 tests pass.
The time was almost all in beforeAll, where bigForge() writes the production-size dataset of about 600k rows (about 25 s when I timed it on its own). The tests only check cache behaviour and that the computation is cut into at least 8 slices. They don't time anything, so they don't need production scale. I added a TENTH size built from PRODUCTION and renamed the first test from "on a production-sized forge" to "on a large forge". No test assertions changed.
The file no longer exercises the flow cache at full production size. Production-size coverage still exists in boot-no-stall, search-production-size and job-slices, but not for this cache specifically. I did not add the file to SLOW in scripts/gate.ts.
To try it: bun test ./test/flow-cache.test.ts.
test/flow-cache.test.ts now builds a tenth-size forge, which brings the file from 23 s to about 4–5 s, under the 20 s file budget. The check's gate failure (a 129-column comment line) is fixed, and I've pushed the fix.
- Cause: nearly all of the 23 s was
beforeAllwriting the production-size forge frombigForge(), about 600k rows. These tests only check cache behaviour and that the computation is cut into at least 8 slices. None of them time anything, so they don't need production scale. I added aTENTHsize built fromPRODUCTIONand renamed the first test from "on a production-sized forge" to "on a large forge". No assertions changed. - Gate fix: the comment above
TENTHis now wrapped under 100 columns.bun scripts/gate.tsexits 0 once the change is committed; it exited 2 on the uncommitted edit, so it evidently reads the committed state. - Checks:
bun x tsc --noEmitprinted nothing and all 9 tests in the file pass. I did not run the full suite; the forge runs it on the push. - Not done: the flow cache is no longer exercised at full production size in this file.
boot-no-stall,search-production-sizeandjob-slicesstill use production-size forges, but not for this cache. I did not add the file toSLOWinscripts/gate.ts.
To try it: bun test ./test/flow-cache.test.ts.
Test-only change: flow-cache.test.ts builds a tenth-size forge (all PRODUCTION counts divide evenly), cutting the file from ~23 s to ~5 s. No assertions changed and the slice-count check (>=8) still holds. No security, API or dependency surface touched. Production-scale coverage of this cache is traded away, which the pull states openly.
test/flow-cache.test.tsnow builds a tenth-size forge, so the file runs in about 5 s here instead of 23 s, under the 20 s file budget.bun scripts/gate.tsandbun x tsc --noEmitwere clean. The file's 9 tests pass.The time was almost all in
beforeAll, wherebigForge()writes the production-size dataset of about 600k rows (about 25 s when I timed it on its own). The tests only check cache behaviour and that the computation is cut into at least 8 slices. They don't time anything, so they don't need production scale. I added aTENTHsize built fromPRODUCTIONand renamed the first test from "on a production-sized forge" to "on a large forge". No test assertions changed.The file no longer exercises the flow cache at full production size. Production-size coverage still exists in
boot-no-stall,search-production-sizeandjob-slices, but not for this cache specifically. I did not add the file toSLOWinscripts/gate.ts.To try it:
bun test ./test/flow-cache.test.ts.Closes #12643