Skip to content

fix(hipdnn): compare model artifact path canonically in descriptor loader test - #14460

Merged
BrianHarrisonAMD merged 3 commits into
developfrom
users/bharriso/windows-temp-path-case
Oct 28, 2026
Merged

BrianHarrisonAMD merged 3 commits into
developfrom
users/bharriso/windows-temp-path-case

Conversation

@BrianHarrisonAMD

Copy link
Copy Markdown
Contributor

Summary

TestDescriptorLoader.AcceptsAModelArtifactInASubdirectoryOfTheDescriptor fails on the Windows nightly because it compares two spellings of the same path. The loader returns the canonical artifact path, which carries the on-disk case (C:\Windows); the test built its expectation from %TEMP%, which may spell it C:\windows or C:\WINDOWS. The test now builds its expectation canonically, matching the loader. The loader itself does not change.

Fixes #14451

Risk Assessment

Risk 1. Test-only change to one assertion; no product code changes.

ASIC Coverage

None. The test is host-only and arch-independent. The failure is Windows-specific, so the Windows nightly hipdnn job (gfx1151, gfx120X-all) is the confirming run.

Testing Summary

  • Linux reproduction: running with TMPDIR set to a symlink makes the literal and canonical paths differ, the same mismatch as the Windows case difference. The old assertion fails this way; the new one passes.

Testing Checklist

  • Descriptor loader tests, symlinked TMPDIR (Linux, -DHIPDNN_ENABLE_KERNEL_INGESTOR=ON) - TMPDIR=<symlink-to-/tmp> hipdnn_plugin_sdk_tests --gtest_filter='TestDescriptorLoader.*' - Status: Passed (173/173)
  • Windows nightly hipdnn tests - gfx1151, gfx120X-all - Status: Pending
  • PR CI - GitHub PR checks - Status: Pending

Technical Changes

  • The expected path uses std::filesystem::weakly_canonical, as the sibling check TestEnginePredictor.AModelPathMustNameAFileInsideTheDescriptorDirectory already does.

…ader test

The UHD parser returns the weakly_canonical artifact path. On Windows that
carries the on-disk case of each component (C:\Windows), while %TEMP% may
spell it differently (C:\windows, C:\WINDOWS), so the absolute() expectation
in AcceptsAModelArtifactInASubdirectoryOfTheDescriptor failed. Build the
expectation with weakly_canonical, matching the parser and the sibling
TestEnginePredictor check.

Fixes #14451
@therock-pr-bot

therock-pr-bot Bot commented Oct 27, 2026 •

Copy link
Copy Markdown

✅ All Checks Passed — Ready for Review

Check Status Details
📝 PR Description ✅ Pass —
⛔ Forbidden Files ✅ Pass —
🧪 Unit Test ✅ Pass PR does not contain code files — Unit Test auto-passed
🔎 pre-commit ✅ Pass —
🚫 Draft PR 🔜 To Be Enabled —
🚩 Feature Flag 🔜 To Be Enabled —
📊 Code Coverage 🔜 To Be Enabled —
🤖 therock-pr-bot ✅ Pass —

🎉 All checks passed! This PR is ready for review.

📖 Need help? See the Policy FAQ for details on every check and how to fix failures.

🙋 Wish to Override Policy?

@therock-pr-bot

Copy link
Copy Markdown

🎉 All checks passed! This PR is ready for review.

@codecov

codecov Bot commented Oct 27, 2026 •

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

Additional details and impacted files
@@             Coverage Diff             @@
##           develop   #14460      +/-   ##
===========================================
- Coverage    77.29%   76.08%   -1.21%     
===========================================
  Files         2889     2886       -3     
  Lines       484830   481477    -3353     
  Branches     73230    72468     -762     
===========================================
- Hits        374716   366288    -8428     
- Misses       85381    90090    +4709     
- Partials     24733    25099     +366     
Flag Coverage Δ *Carryforward flag
TensileLite-CPP 50.80% <ø> (-0.01%) ⬇️ Carriedforward from a11ca45
TensileLite-Unit 84.10% <ø> (+0.08%) ⬆️ Carriedforward from a11ca45
hipBLAS 90.62% <ø> (ø) Carriedforward from a11ca45
hipBLASLt 66.14% <ø> (-2.82%) ⬇️ Carriedforward from a11ca45
hipCUB 82.68% <ø> (ø) Carriedforward from a11ca45
hipDNN 86.85% <ø> (-<0.01%) ⬇️
hipFFT 38.02% <ø> (ø) Carriedforward from a11ca45
hipRAND 76.12% <ø> (ø) Carriedforward from a11ca45
hipSOLVER 69.67% <ø> (ø) Carriedforward from a11ca45
hipSPARSE 87.03% <ø> (ø) Carriedforward from a11ca45
rocBLAS 68.73% <ø> (-0.23%) ⬇️ Carriedforward from a11ca45
rocFFT 47.34% <ø> (ø) Carriedforward from a11ca45
rocRAND 57.31% <ø> (ø) Carriedforward from a11ca45
rocSOLVER 76.74% <ø> (ø) Carriedforward from a11ca45
rocSPARSE 77.16% <ø> (-6.38%) ⬇️ Carriedforward from a11ca45
rocThrust 93.15% <ø> (ø) Carriedforward from a11ca45

*This pull request uses carry forward flags. Click here to find out more.
see 253 files with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

@BrianHarrisonAMD
BrianHarrisonAMD merged commit 413ce9b into develop Oct 28, 2026
180 of 183 checks passed
@BrianHarrisonAMD
BrianHarrisonAMD deleted the users/bharriso/windows-temp-path-case branch October 28, 2026 22:50
@assistant-librarian

Copy link
Copy Markdown
Contributor

TheRock Submodule Bump Activity

Newest first

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

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Issue] Windows nightly: TestDescriptorLoader.AcceptsAModelArtifactInASubdirectoryOfTheDescriptor fails on Windows temp-path case mismatch

2 participants