bst1: Support Python 3.11#1785
Conversation
nanonyme
force-pushed
the
nanonyme/python3.11-bst1
branch
from
2ecf22e to
e486c27
Compare
|
Why is this still a draft? |
|
@mcatanzaro tests don't pass and it's not directly related to Python 3.11 |
|
I am not getting a pipeline, dunno why. |
nanonyme
force-pushed
the
nanonyme/python3.11-bst1
branch
from
d540307 to
fe9835f
Compare
|
I'm not entirely sure if tests hang on Fedora 36 and 37 but at least they're massively slower. |
nanonyme
force-pushed
the
nanonyme/python3.11-bst1
branch
from
02726a2 to
99c6646
Compare
|
Looks like there are failing tests. The failing tests are failing really slowly. I'm trying to make tests abort after first failure so we see real output. |
|
Well, that was not overly useful. I got it to actually end and we're maybe a bit closer to understanding why the test is so slow https://github.com/apache/buildstream/actions/runs/3688689726/jobs/6243779599 but the traceback is mostly gibberish and there is no console output |
nanonyme
force-pushed
the
nanonyme/python3.11-bst1
branch
5 times, most recently
from
cb0d0c6 to
ff24375
Compare
|
Hi @gtristan, can you look at this please? buildstream 1 is really super broken currently, and for a very long time now (it's not like python 3.11 is new anymore). |
|
Note that this is about CI and testing on Python 3.11. This has no functional changes. |
|
There are unreleased fixes for Python3.11 in bst-1 branch. But without CI it's hard to say whether further things are broken. |
|
Er, why not release them then? This has been blocking a lot of stuff for a long time. (I agree that CI is very important, but it's not urgent.) |
|
It would be nice if someone pip installed bst1 from this repo who actually has Python3.11 to give it at least some lightweight manual testing. |
|
@mcatanzaro @nanonyme I think that is what is being done in #1757 (It hangs forever) |
|
@jjardon I am not taking about CI. Just install it and use it. It will not hang. See if there are further regressions. |
|
If it seems usable, then we tag a release and worry about CI later. |
|
The hangs are related to special way bst event loop is used for unittest purposes in same process restarting it over and over again. They are not indicative of real problems. |
|
@jjardon it also does not hang "forever". Asyncio gets into bad state and times out with 30 minutes at a time. It just times out dozens of times in a loop so it looks like forever. |
You don't have access to an F37 toolbx container? That should take maybe two minutes to create. Whatever. I installed this branch and am building gnome-build-meta with it now. Will let you know how it goes.... |
Well it finished successfully, but I think it pulled everything from the artifact cache without actually building anything. Want me to test anything else in particular? |
|
I think that's strong indication we should wrap a quick release. If you can build something (doesn't matter what, you can force rebuild by setting environment variable in some element), that would be even better. |
|
I created #1809 anyhow. I would like to have CI actually work before advertising Python 3.11 support in metadata. But this should hopefully unblock downstreams. |
Did this to force building core/epiphany.bst. It worked. |
And remove soon to be deprecated fedora 35
This is the python version in fedora 37
The older version doesn't support Python 3.11
This is first upstream release that advertises Python 3.11 support
nanonyme
force-pushed
the
nanonyme/python3.11-bst1
branch
from
ff24375 to
9df051c
Compare
|
This is replaced by #1827 |
4 participants
Fixes #1758