How should source-only deepagents-evals CLI bugs be reported? #6570
Unanswered
Lucien (dakjdakd)
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
I found a reproducible exit-code bug in the
deepagents-evals runCLI on monorepomainat0c756c4f8.libs/evals/tests/evals/pytest_reporter.py::pytest_sessionfinish()intentionally resets pytest exit status 1 to 0 after failed tests so aggregation can finish.libs/evals/deepagents_evals/cli.py::_cmd_run()checks only that subprocess status, so a run whose report contains{"failed": 1}exits 0. Automated callers therefore see success for a failed evaluation.trialsandaggregatealready inspect report failure counts.A network-free CLI reproduction:
Observed on
main:0; expected:1when the report records a failed evaluation. A separate check through the real reporter hooks producedpytest_exit=0andreport_failed=1. Related #3221 introduced the unified CLI, and #2932 addresses the no-tests-run exit, not this path.I would report this with the Bug Report template, but its required checklist says the bug must reproduce with the latest released version. The published
deepagents-evals==0.0.1package is a placeholder; this CLI exists only in monorepo source, so I cannot truthfully check that item. What is the preferred route for a source-only evals bug? If this discussion can serve as the approved contribution link, I can provide a focused fix and regression tests once a maintainer approves the work and assigns me (@dakjdakd). Otherwise I can file an Issue through the route you recommend.All reactions