Repository navigation
test(tinyaml): confirm a slope over 1.8 at 4x on 16x before failing - #31
Conversation
A linear reader came out at 1.87 at 4 times the input on CI (Node 22, reading empty documents with tinyaml/js-yaml), and the early exit failed it there. Now a slope over 1.8 at 4x reads 16x once more and that slope decides, so a quadratic reader still fails soon.
Codex Review Summary: ❌ FailedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes — Reviewed the scaling-test helper change in packages/tinyaml/test/fixtures.ts and its callers.
- Confirm before failing — A 4x slope above 1.8 now triggers a 16x measurement, whose slope determines the result instead of the intermediate measurement.
- Preserve the normal path — The existing 16x retry path, sampling budgets,
LINEARthreshold, and stack-trace restoration remain unchanged; comments document the CI false positive.
Validation: All 360 tests across fixtures.test.ts, scaling.test.ts, root.test.ts, and linter.test.ts passed. A deterministic CPU-clock probe also confirmed that a noisy 4x slope of 1.87 resolves to 1 for linear 16x work, while quadratic work still returns 2.
gpt-6.1-sol | 𝕏
commit: |

The last
mainrun failed on Node 22:reading empty documents > takes linear time in tinyaml/js-yamlmeasured a slope of 1.87 (run). That reader is linear. Locally it comes out at 1.04 to 1.35 on Node 22 and 24, and at 0.95 to 1.27 on one core shared with a busy loop. The same shape failed earlier for js-yaml-v5, std and yamljs too.slope()reads 4 times the input on the way and stops when the slope there is over 1.8. That gate decided on its own, and at 4x the base run is short, so a pause or the 0.1 ms CPU clock moves it a lot.Now a slope over 1.8 at 4x reads 16 times the input once more, and that slope decides. A linear reader passes there. A quadratic one still fails without the three retry rounds: a toy quadratic reader fails at 2.10. The usual path is unchanged when 4x stays under 1.8.
I couldn't reproduce the 1.87 locally, so this is checked against the CI log and local runs, not the failing runner. Node 24.21.0 and 22, Linux.