Closed kiukchung closed 10 months ago
will graduate from draft commit after I validate GHA and check codecov reports
codecov metrics + flags look good
Patch coverage has no change and project coverage change: +6.79%
:tada:
Comparison is base (
a465816
) 76.82% compared to head (742e71c
) 83.62%. Report is 3 commits behind head on main.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
This PR does a few things:
keras_core/legacy
from coverage metrics since this module there for BC reasonskeras_core
will have higher coverage thankeras_core-jax
), so this flag should be used to track trends rather than impose thresholds.coverage
andpytest
configuration fromsetup.cfg
topyproject.toml
to keep all project configs in toml vs cfg (unfortuantelyflake8
configs need to stay insetup.cfg
since flake8 doesn't supportpyproject.toml
yet: https://github.com/PyCQA/flake8/issues/234)fail_ci_if_error: true
to codecov upload steps to fail the CI (and force the author to retry the workflow) on transient codecov upload failures of the formThis happens when we hit codecov API rate limits since we don't use
CODECOV_TOKEN
(see https://github.com/codecov/feedback/issues/126). We could circumvent the issue by adding the token as a secret but GH secrets are not visible from forks, implying that contributors will not be able to get coverage diffs for their PRs. To allow this, the only option seems to be to hardcode the token (which seems ok since the token can be setup with limited permissions) but hardcoding tokens is just sketchy...