Repository navigation
test(vx): deal the Mac's shards by the Mac's own times - #5330
Merged
Merged
Conversation
The Mac platform shards are dealt by weights measured on Linux and run 57 to 88 s. A successful task's output is not in a worker's log, so each Mac shard prints its files' times and fails. Read the times, close the PR. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W6xuYXmaeAXEqaqA1Z6RsZ
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W6xuYXmaeAXEqaqA1Z6RsZ
The probe printed each platform file's time on a Mac and failed on purpose; its times are read. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W6xuYXmaeAXEqaqA1Z6RsZ
A Mac skips the Linux sandbox's tests and runs its own, so the Linux table dealt the six platform shards 57 to 84 s of work, the same order on two runs (main 93d6a41: 57 to 88 s). The Mac work is 630 to 660 core-seconds on three Macs, as long as the Linux work now, and its shards run first: the longest one sets when the rest can start. tests/platform-shard-weights.json holds each platform file's time on CI's Macs, read off a probe run (#5330's first commits: a passing task's output is in no log). By those times each shard carries 70.5 s. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W6xuYXmaeAXEqaqA1Z6RsZ
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The six Mac test shards carry the same load
Before: the Mac platform shards were dealt by weights measured on Linux. A Mac skips the Linux sandbox's tests and runs its own, so the six shards ran 57 to 84 s (88, 83, 75, 80, 69, 57 s on main 93d6a41; 84, 71, 74, 76, 62, 57 s on this PR's probe run: the same order). The Mac work is 630 to 660 core-seconds on three Macs, as long as the Linux work now, and the shards run first: the longest sets when the rest can start.
After:
tests/platform-shard-weights.jsonholds each platform file's time on CI's Macs, and--platformdeals by it. By those times each shard carries 70.5 s. A platform file the table does not know keeps its Linux time.How it works
This PR's first two commits were a probe: a passing task's output is in no log, so each Mac shard printed its JUnit file times and failed on purpose (run 44234002954, all 122 files; each shard's files summed to the shard's own time to the second). The third commit takes the probe out; the last one adds the table.
readWeights(suite, platform)lays the Mac table over the Linux one.Breaking changes
None.
Proof
tests/shard-partition.test.ts"deals the Mac's shards by the Mac's own times": the deal's loads by the Mac's times are within 5 %. Without the dealer change it fails at 1.47.🤖 Generated with Claude Code
https://claude.ai/code/session_01W6xuYXmaeAXEqaqA1Z6RsZ