Repository navigation
test: stabilize native macOS alias and coverage checks - #581
Conversation
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Compare only verified setup-python shortcuts across native architectures. Observe cache, output, and descendant lease waits before releasing their owners so LLVM instantiation coverage does not depend on scheduling. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Keep the controlled runs as immutable evidence without adding a permanent diagnostic workflow or changing the native coverage gate. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Performance Report (Linux)Result: ✅ Within regression budgets
|
Test Coverage Report (Linux)Result: ✅ Within regression budget
|
Performance Report (macOS ARM64)Result: ✅ Within regression budgets
|
Test Coverage Report (Windows)Result: ✅ Within regression budget
|
Performance Report (Windows)Result: ✅ Within regression budgets
|
Performance Report (macOS)Result: ✅ Within regression budgets
|
|
🔒 Automated review in progress — Rich Chiodo (@rchiodo) is auto-reviewing this PR. |
|
Reviewed commit: Result: 🔴 Verification detailsVerification: Automated verification is best effort. The PR author remains responsible for testing. Summary: [unavailable] Container verification could not start and local execution was not authorized for this PR HEAD: Podman CLI is not installed or is not on PATH. Test runs: none recorded. |
Rich Chiodo (rchiodo)
left a comment
There was a problem hiding this comment.
Approved via Review Center.
Summary
Independent follow-up to merged #579, based on
5ac67ce6dc78dd24c2d0e766bd9b07c1f161721a. No production locator, production wait/shutdown semantics, coverage comparator, workflow, tolerance, workload, or performance lane is changed.MacPythonOrgaliases matching/Users/runner/hostedtoolcache/Python/<major.minor.patch>/{x64,arm64}/bin/pythonqualify, and each must resolve to that environment's executable before exclusion on either side. Different/missing targets fail; ordinary aliases and metadata comparison remain intact. Focused cases cover both architectures, unrelated/malformed paths, non-ASCII/non-UTF8 components, and different/broken targets.cfg(test)-only and bounded; the existing owner guard still wakes waiters on unwind.Controlled native evidence
The merged-main Intel failure independently reproduces the extra x64 toolcache alias: run 43978821541 / job 133689660257.
Before changing coverage behavior, run 43981161006, attempt 1 measured the same unchanged main commit in two separate source/target trees in predeclared A/B/B/A order. Artifact 14283815268,
macos-aa-43981161006-1, retains full LLVM JSON with function/instantiation regions, LCOV, repeated exports, raw profiles, contexts, and isolated child proof.The unchanged same-path B1/B2 comparison fails at -0.013pp. Re-exporting the same profiles is byte-identical in every pass. LLVM regions identify cache waiter paths and
writer_loop<FailingWriter>scheduling-dependent hits; a positive union ofDAlines hides the latter instantiation difference. This establishes runtime path variability in this experiment, not a compiler/exporter defect or a revision-dependent regression. Original artifact 14275796507 separately identifies the descendant lease's three missed pending-poll lines.Patched run 43983235000, attempt 1, artifact 14284656224,
macos-aa-43983235000-1, passes all four predeclared raw comparisons. Cache and output summaries are identical across its four passes. Residual spread is one test-helper line (0.0044pp), retained without normalization and below the unchanged 0.01pp gate. Function totals are constant within each experiment: main 1826/2205; patched 1829/2208.Both experiments use Rust 1.100.0 / LLVM 23.1.5 / cargo-llvm-cov 0.9.1 on native ARM64 macOS. Every isolated subprocess proof remains idle=0 / info=1 at handler, transport, and writer. Temporary dispatch-only experiment changes were removed completely from the final diff.
Review and validation
cargo fmt --allandcargo clippy --all -- -D warnings, additional changed-package test-target Clippy, and 17 workflow-contract tests passed.Hosted PR validation at
cc3379ba77b7e5221b72824f9d8a560d4802424cAll six repository workflows passed on their first PR attempt, as did the external policy check:
133708333703, ARM64 job133708333963verify_validity_of_discovered_envs ... ok, its suite 6 passed / 0 failedAll commits are locally GPG-signed and GitHub-verified. No merge or auto-merge is enabled; no merge is requested or authorized.