Skip to content

test(allocation): set the InsertBatch ceiling from runner measurements (915 B/row) - #504

Merged
MPCoreDeveloper merged 1 commit into
masterfrom
fix/allocation-ceiling
Nov 12, 2026
Merged

MPCoreDeveloper merged 1 commit into
masterfrom
fix/allocation-ceiling

Conversation

@MPCoreDeveloper

Copy link
Copy Markdown
Owner

master merged the promotion at 4c078b8a, which carried the per-OS allocation ceiling — OperatingSystem.IsWindows() ? 900 : 915. The Windows figure was recorded on a developer machine (894 B/row), but the GitHub windows-latest runner measures 902 B/row for the same commit, so the Windows job of Build and Test fails on master and the pack → publish jobs are skipped (needs: build), which blocks the 2.1.3 publication.

This PR replaces that with the ceiling measured on the runners, which was pushed five minutes after the merge (26fdcfde / the single commit here):

Environment Measured Ceiling
developer box, win-x64 894 B/row (1.787.608 B) —
GitHub windows-latest 902 B/row (1.803.560 B) 915
GitHub ubuntu-latest 903 B/row (1.806.872 B) 915
GitHub macos-latest (arm64) 903 B/row (1.806.856 B) 915

915 is above the highest runner figure and still below the pre-hoist 919 B/row (~927 on a runner), so reintroducing the allocation this canary was written to catch still fails on every platform. The test comment records all four measurements and the guardrail: absolute-bytes canaries must be set from the runner measurement, never a developer machine.

Verified locally: the canary passes on win-x64 with 915 (894 measured). Ubuntu and macOS already passed at 915 in the previous run of the promotion PR — only Windows was above the old 900 ceiling.

…s (915 B/row)

The per-OS ceiling was still wrong: the GitHub windows-latest runner measures 902 B/row while this developer box measures 894 on the same commit, and both Unix runners measure 903. So the spread is the environment's GC/JIT accounting, not the OS and not a code path - an absolute-bytes canary set on a developer machine fails every runner.

One ceiling now covers all four environments: 915 B/row, above the highest runner figure (903) and below the pre-hoist 919 (about 927 on a runner), so reintroducing the allocation this canary was written to catch still fails everywhere. Measurements recorded in the test: dev box win-x64 894 (1.787.608 bytes), runner win-x64 902 (1.803.560), runner linux-x64 903 (1.806.872), runner osx-arm64 903 (1.806.856). Verified locally: the canary passes on win-x64 with the new ceiling.
@codecov

codecov Bot commented Nov 12, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@MPCoreDeveloper
MPCoreDeveloper merged commit d714ef9 into master Nov 12, 2026
15 checks passed
@MPCoreDeveloper
MPCoreDeveloper deleted the fix/allocation-ceiling branch November 12, 2026 21:44
@sonarqubecloud

Copy link
Copy Markdown

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant