[CI] Restrict LTS coverage tracking to the checkout - #645
Merged
Merged
Conversation
`--code-coverage=user` covers every non-Base package, so each of the ~45 parallel test workers writes a .cov file for every dependency source file (GPUCompiler, LLVM, GPUArrays, ...) into the CI depot on Lustre. That slows worker start-up and shutdown enough to trip Malt's hard-coded 30 s connect and 20 s exit timeouts, failing otherwise green runs (pipelines 65611 on 1.10 and 65820 on 1.13). None of those files are consumed: the lcov step only processes src/, lib/ and examples/. Track `@$CI_PROJECT_DIR` instead; Base.julia_cmd() forwards the tracked path to the workers.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #645 +/- ##
==========================================
- Coverage 81.61% 80.59% -1.02%
==========================================
Files 50 54 +4
Lines 3611 3855 +244
==========================================
+ Hits 2947 3107 +160
- Misses 664 748 +84 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 ALCF LTS job ran the tests with
--code-coverage=user, which also covers every dependency. Each of the ~45 parallel test workers writes a.covfile for every source file it loaded (GPUCompiler, LLVM, GPUArrays, KernelAbstractions, ...) into the CI depot on Lustre. The two CI depots had piled up ~43k and ~51k such files (~1 GB each).That I/O makes workers slow to start and to exit, which trips Malt's hard-coded timeouts and fails runs where every test passed:
d2e06ce,test:lts:1.10:Worker process exited before we could connect … caused by: Timeout(30 s connect). 43/45 tests passed and 2 never got a worker.f110df6,test:lts:1.13: all 45 tests passed, thenHOST: Worker did not exit after 20.0 seconds.None of those depot
.covfiles are used, because the lcov step only processessrc/,lib/andexamples/. This PR switches to--code-coverage=@$CI_PROJECT_DIR.Checked locally with Julia 1.10:
Base.julia_cmd()passes the tracked path on to child processes (code_coverage == 3with the checkout as tracked path). Coverage files appear undersrc/andlib/, and none under the depot'spackages/.🤖 Generated with Claude Code