fix: re-raise on test results CSV save failure

[rgutzen]

Problem

process_single_test was failing with a confusing pandas.errors.EmptyDataError: No columns to parse from file when reading test_outputs.csv, even though the upstream test rule had reported success.

Root cause

In dynvision/runtime/test_model.py::TestingOrchestrator.save_results(), the block that writes test_outputs.csv catches all exceptions and only logs them:

except Exception as e:
    logger.error(f"Failed to save test results: {e}")

If model.storage.get_dataframe() or to_csv() fails partway (disk error, OOM while formatting, quota, interrupted write, etc.), the exception is swallowed and save_results() returns normally. The shell command then exits 0 with a missing/empty/truncated test_outputs.csv on disk. Snakemake treats the output as successfully produced and never reruns the job — the corrupted file then silently propagates to downstream processing steps, which fail later with an unrelated-looking pandas error.

This is inconsistent with the response-tensor save path immediately below it in the same method, which already re-raises on failure specifically so Snakemake can detect it via a missing output file.

Fix

Re-raise after logging in the CSV-save except block, mirroring the existing response-save error handling, so a failed write causes the Snakemake job itself to fail (and be rerun) rather than leaving a corrupt output silently marked complete.

Testing

• ast.parse syntax check on the modified file.
• No existing unit test directly covers save_results(); behavior verified by code inspection and by reproducing the downstream symptom (EmptyDataError from an empty test_outputs.csv) against the described failure mode.

[Copilot]

🟢 Approval recommended

The change is minimal, consistent with the existing response-save error handling in the same method, and directly addresses the described root cause by ensuring Snakemake observes failed writes as job failures.

Pull request overview

This PR fixes a failure-mode in the testing runtime where writing test_outputs.csv could fail silently, leaving behind a missing/empty/truncated CSV that Snakemake still treats as a successful output, causing confusing downstream errors when the corrupted file is read later.

Changes:

• Re-raise exceptions encountered while saving the test results CSV (after logging), so the Snakemake job fails and can be rerun rather than producing a corrupt output artifact.

File summaries

File	Description
dynvision/runtime/test_model.py	Re-raises on test_outputs.csv write failures to prevent silently “successful” runs with missing/corrupt outputs.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.