Skip to content

fix(scheduler): stop the live experiment shadowing the failed_jobs count - #314

Closed
bpiwowar wants to merge 1 commit into
masterfrom
fix/failed-jobs-shadowing
Closed

bpiwowar wants to merge 1 commit into
masterfrom
fix/failed-jobs-shadowing

Conversation

@bpiwowar

@bpiwowar bpiwowar commented Nov 5, 2026

Copy link
Copy Markdown
Collaborator

Problem

experiment.failed_jobs() (returning the (failed, not_run) job lists) overrides the BaseExperiment.failed_jobs property (the failed-job count). For a running experiment:

  • the web UI fails to list experiments: Error handling message experiments: '>' not supported between instances of 'method' and 'int' (from status_category, self.failed_jobs > 0);
  • the remote server (remote/server.py) sends a bound method as failed_jobs.

Seen with experimaestro run-experiment --web on the demo, when loading the web page.

Fix

Rename the method to failed_jobs_by_reason() (callers: failure_summary, _report_failed_jobs, test_tasks), and add a test checking that the live experiment does not override the failed_jobs property.

🤖 Generated with Claude Code

`experiment.failed_jobs()` (the failed / not-run job lists) overrode the
`BaseExperiment.failed_jobs` count property, so the web UI failed to list
a running experiment ("'>' not supported between instances of 'method'
and 'int'" in status_category) and the remote server sent a bound method.
The method is renamed `failed_jobs_by_reason()`.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@bpiwowar

bpiwowar commented Nov 5, 2026

Copy link
Copy Markdown
Collaborator Author

Pushed directly to master as bb28fee.

@bpiwowar bpiwowar closed this Nov 5, 2026
@bpiwowar
bpiwowar deleted the fix/failed-jobs-shadowing branch November 5, 2026 10:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant