Fix speccpu2017 processor to find CSVs in nested result directories

[sayalibhavsar]

Summary

• The speccpu2017-wrapper v2.6 produces archives where CSV results are nested inside results_speccpu_*/result/ rather than directly in result/
• The processor searched one level deep with glob("*.csv"), causing all SPEC CPU 2017 results to be silently empty, an empty document was indexed to OpenSearch with no suite data
• Switched from glob() to rglob() so the processor discovers CSVs regardless of directory nesting depth

Archive structure produced by wrapper v2.6

results_speccpu2017.zip
  └── results_speccpu2017_virtual-guest.tar
        └── speccpu2017_{timestamp}/                          ← extracted_path
              ├── meta_data.yml
              ├── version
              └── results_speccpu_virtual-guest_{timestamp}/  ← CSVs are here
                    └── result/
                          ├── CPU2017.003.intrate.refrate.results.csv
                          └── CPU2017.004.fprate.refrate.results.csv

The processor searched extracted_path/ and extracted_path/result/ but never extracted_path/results_speccpu_*/result/.

Test plan

• New test test_speccpu2017_csv_in_nested_results_subdir covers the nested directory case with both intrate and fprate CSVs
• All 217 tests pass (was 216 before, +1 new test)
• Verified against real artifacts from Azure CPT pipeline job 53844821 : confirmed valid intrate (10 benchmarks) and fprate (13 benchmarks) data exists but was not being found
  fixes speccpu2017_processor fails to find CSV results - incorrect directory search depth #21

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR modifies SpecCPU2017Processor to discover SPEC CPU 2017 suite CSV files using recursive directory traversal instead of non-recursive scan, enabling the processor to find timestamp CSVs in nested result subdirectories. A test validates the behavior with the wrapper v2.6+ directory layout.

Changes

Recursive CSV Discovery in Nested Results

Layer / File(s)	Summary
Recursive CSV file discovery in fallback path src/chronicler/processors/speccpu2017_processor.py	SpecCPU2017Processor.parse_runs replaces glob("*.csv") with rglob("*.csv") to recursively search the chosen result directory for intrate and fprate timestamp CSVs.
Test nested CSV discovery tests/test_speccpu2017_timestamps.py	New test test_speccpu2017_csv_in_nested_results_subdir verifies the processor discovers and parses SPEC CPU 2017 intrate and fprate CSVs located under nested results_speccpu_*/result/ subdirectories.

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'Fix speccpu2017 processor to find CSVs in nested result directories' clearly and concisely summarizes the main change: switching from glob() to rglob() to discover CSV files at any nesting depth.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description check	✅ Passed	The PR description clearly explains the issue with speccpu2017-wrapper v2.6 producing nested CSV directories, the specific problem with glob() vs rglob(), and includes the exact archive structure and test coverage.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[grdumas]

LGTM