cc @methane @pablogsal @zware: Do you think that 20 minutes per test is enough for everyone?

If someone has issues with the new timeout, TESTTIMEOUT can be overriden when running make: make TESTTIMEOUT=3600 (1 hour).

In regrtest --pgo, the slowest test should be asyncio, and it takes around 5 minutes.

If 20 min is an issue, maybe we can increase the default to 30 minutes?

---

It's the first time that I see a hang while on a PGO build. So another option is to consider that no timeout is better by default (reject this PR).

I would be fine with that but we should certainly do something about test_asyncio

I would be fine with that but we should certainly do something about test_asyncio

I don't think that we should do anything for "aarch64 RHEL7 LTO + PGO 3.7", since 3.7 branch no longer accept bug fixes.

I'm thinking of removing that config from the aarch64 RHEL7 buildbots, as it's the only one that disconnects during a PGO build when test_asyncio runs.

I'm thinking of removing that config from the aarch64 RHEL7 buildbots, as it's the only one that disconnects during a PGO build when test_asyncio runs.

I created python/buildmaster-config#228 for that.

Do you think that 20 minutes per test is enough for everyone?

I think so. And I think tests takes longer than 20min should not be run by default.

The timeout can cause some issues, so I prefer to not backport it. https://bugs.python.org/issue43031 is not a major issue, it only impacted Python 3.7. I added the timeout in Python 3.10, it is enough for now.

Thanks for your review @pablogsal and @methane.