[saime428]

Summary

• Set format_version=1 after copying the source attributes.
• Add parametrized tests for legacy files with no version attribute and with an explicit version 0. Both cases check the output version and the root pose after loading on CPU.

Why

The converter set format_version=1 before recursively copying the legacy file's attributes. If the source had format_version=0, that value overwrote the new marker. Loading the converted file then treated its XYZW poses as WXYZ and converted them a second time.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Validation

• New tests against the unchanged implementation at 259e1a8: 1 failed, 1 passed. The explicit-version-0 case failed because the output version was still 0.
• Full test_hdf5_dataset_file_handler.py on the final implementation with ISAACLAB_TEST_DEVICES=100: 6 passed.
• uv run --no-sync isaaclab --format: all applicable checks passed across the repository, using Git Bash on Windows.
• git diff --cached --check before committing.

Notes for reviewers

• Tested on Windows with Python 3.12.13, h5py 3.15.0, NumPy 2.5.4, and PyTorch 2.11.0+cpu. The package was installed editable from this checkout in an isolated environment with the dependencies needed by this test file.
• Both test runs emitted the same 35 torch.jit.script deprecation warnings from unchanged math utilities.
• GPU, Isaac Sim, and the full project test suite were not run. The changed conversion and regression use CPU file I/O and do not launch a simulator.

Checklist

• Read the contribution guidelines.
• Run the repository pre-commit checks.
• Validate the changed behavior with focused coverage and apply the test-audit criteria.
• Add a changelog fragment for the touched package.
• Add my name to CONTRIBUTORS.md.
• Confirm no new warnings in the focused tests.

No documentation changes are needed for this fix.

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge.

Summary

Fixes legacy HDF5 conversion by setting format_version after copying source attributes.

• Converted HDF5 datasets keep the version that matches their poses.

No actionable issues found. Tests were not run during this review.

_{Reviews (1) · Last reviewed commit: "Fix the format version after converting ..." · Reviewed by Greptile}

[isaaclab-review-bot]

Isaac Lab Review Bot

All reviewed code categories look good.

• PR description: The title and body match the diff, explain the failure path, and provide focused baseline and final validation; no performance evidence is needed.
• Design and architecture: Architecture looks good.
• Scope and over-engineering: No over-engineering issues.
• API: API changes look good.
• Compatibility and deprecation: No breaking changes identified.
• Implementation: Implementation looks good.
• Style consistency: Style consistency looks good.
• Test quality: Test quality looks good.

No blocking issues. No inline findings.

Automated review; human maintainers own approval decisions.

[AntoineRichard]

Thanks a lot @saime428 for the fix!

[AntoineRichard]

run-ci