Python 3.10 Wheels #152
Python 3.10 Wheels #152
Conversation
|
Well this looks working now. I think the builds are flaking out again, anyone(with write access) want to restart them(I could close/reopen but that wastes a ton of build minutes)? |
|
Ok this should be green now for travis also. Have also confirmed that the other 32 bit builds aren't flaking out and seems to also be an actual failure on master, so I'll upgrade/fix them later in another PR. travis failed because of a flakeout on py 3.7, the py 3.10 build on their is fine. Let's just restart all non 32-bit flaky jobs just to be sure. |
.travis.yml
Outdated
| @@ -60,7 +69,15 @@ script: | |||
| - source multibuild/travis_steps.sh | |||
| - source extra_functions.sh | |||
| - setup_test_venv | |||
| - install_run $PLAT | |||
| # pandas-dev/pandas#41935, we get weird numpy | |||
| # failures on python 3.10 | |||
There was a problem hiding this comment.
This is not failing on pandas ci?
There was a problem hiding this comment.
It was, but we just made the CI allowed to fail. The failed tests are visible when you look at the logs of any Python 3.10 run I believe.
There was a problem hiding this comment.
we just made the CI allowed to fail
maybe the failing tests should be XFAILed instead?
There was a problem hiding this comment.
There are a lot of flaky ones unfortunately, which is why the build is just allowed to fail.
.travis.yml
Outdated
| - if [ "$MB_PYTHON_VERSION" == "3.10" ]; then | ||
| set +e; | ||
| install_run $PLAT; | ||
| exit 0; |
There was a problem hiding this comment.
how many tests are failing? can we just update function run_tests in config.sh to skip those tests? we do this for others.
There was a problem hiding this comment.
I'm actually not too sure. The failures here are different than those in the main repo. It seems though like its the same group of 23 tests failing here, though. I'll try to update when I have time later.
|
@simonjayhawkins This should be ready now. Not removing the other code that makes CI pass since we could have different failing tests on master(this PR builds on 1.3.2). I'll try to follow up on this. (Travis build canceled, but the last one passed. The last commit is just modifying a comment). |
Thanks @lithomas1 . I would prefer not to have the extra code but also want to see the 3.10 wheels built on nightly sooner. since nightly is built from master is it worth temporarily changing |
There was a problem hiding this comment.
Thanks @lithomas1 lgtm
NTERNALERROR> assert not 'tests/base/test_unique.py::test_unique_bad_unicode[idx_or_series_w_bad_unicode1]' seen before. restarted to be sure. ping on green
|
@simonjayhawkins As green as it gets, I guess. 3.7 linux flaked out again with a different but still flaky error. |
|
Thanks @lithomas1 |
xref pandas-dev/pandas#42946
(this also needs a whatsnew, pls don't close the issue)