Skip testing with PyTorch 1.7 and Python 3.9 on Ubuntu

[akihironitta]

What does this PR do?

Fixes #11184.

Skips testing with PyTorch 1.7 and Python 3.9 on Ubuntu due to a known issue with the setting: #11217 (comment)

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• [n/a] Did you make sure to update the documentation with your changes? (if necessary)
• [n/a] Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• [n/a] Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

cc @carmocca @akihironitta @Borda

[akihironitta]

wip

[carmocca]

Looks like it's not the coverage as it still segfaulted.

Reading online, this could be useful

https://docs.python.org/3/library/faulthandler.html

python -X faulthandler -m pytest tests ...

[carmocca]

So the problem was reported already but the fix only made it to 1.8, we'll need to exclude this job parametrization

[akihironitta]

@carmocca Thank you very much for taking your time and investigating this! 🎉

So the problem was reported already but the fix only made it to 1.8, we'll need to exclude this job parametrization

I had a look at the GitHub issue before and had it in my mind, but simply running gdb didn't provide such info (i.e. pybind11::gil_scoped_acquire...). I'm curious, for future debugging, about how you've got to see the above core dump. I'm assuming you've done via ssh to the session, correct? Confirmed so via ssh using tmate.

[akihironitta]

Just for the record, this issue seems to have been resolved in #50998 which is included in v1.8.0 as seen in their release note: https://github.com/pytorch/pytorch/releases/tag/v1.8.0

[Borda]

btw, I would not run any oldest for py3.9

- {python-version: "3.9", requires: "oldest"}

[akihironitta]

@Borda Could I ask why you wouldn't? This problem shouldn't apply to other oldest settings.

[Borda]

use release: "stable" is redundant since we do not have other release values then stable now
and in general, if someone is using almost the latest python I don't think he would assist on the oldest packages...

[akihironitta]

use release: "stable" is redundant since we do not have other release values then stable now

👍

in general, if someone is using almost the latest python I don't think he would assist on the oldest packages...

I see your point. I've just created an issue #11367 to gain some visibility on this.