otisolsen70 has quit [Remote host closed the connection]
otisolsen70 has joined #pypy
Atque has quit [Quit: ...]
<mattip>
when I copy Lib/* from CPython3.9x on top of py3.9, we seem to be missing _weakref._remove_dead_weakref
<mattip>
I wonder what I did when vendoring the stdlib
<mattip>
ahh, there is a pure-python implementation for py3.9
<mattip>
... and it calls _delitem_if_value_is which has a scary comment about a race condition with multiprocessing
<mattip>
... but that function has been in __pypy__ for quite a few years now
<mattip>
... well, it doesn't matter since that codepath is not hit in the hanging compileall
Dejan has joined #pypy
jacob22_ has quit [Ping timeout: 240 seconds]
slav0nic has joined #pypy
jacob22_ has joined #pypy
<mattip>
cfbolz: the PR that has commit 1ac6e379297cc1cf8acf6c1b011fccc7b3da2cbe was part of 3.9.0a4, right?
<mattip>
cfbolz: if I use multiprocessing from 3.9.0a1, I still get a hang
<mattip>
(only copying multiprocessing)
<cfbolz>
mattip: you need to copy concurrent futures too
<mattip>
but then I get an error in multiprocessing/connection _check_closed "handle is closed" at shutdown
<cfbolz>
mattip: right, I found that too
<cfbolz>
mattip: for me, c12375aa0b838d34067efa3f1b9a1fbc632d0413 doesn't hange
<mattip>
right, can replicate c12375aa0b838d34067efa3f1b9a1fbc632d0413 does not hang
<cfbolz>
mattip: and its child commit hangs
<cfbolz>
mattip: but it could still be a pypy bug, because what the commit does is make the forks happen later
<cfbolz>
that means it's possible that some object that is created between the non-hanging time the fork happens and the hanging time is responsible for the problem after forking
jacob22 has joined #pypy
jacob22_ has quit [Ping timeout: 250 seconds]
<mattip>
I also get that 1ac6e379297cc1cf8acf6c1b011fccc7b3da2cbe is the commit that after that compileall hangs
<mattip>
that commit touches concurrent/futures/process.py and adds a threading.Semaphore(0)
Atque has joined #pypy
lritter has joined #pypy
Atque has quit [Quit: ...]
<mattip>
we could revert the lazy process creation for PyPy which avoids hanging compileall,
<mattip>
but indeed it would be better to find the real problem
greedom has joined #pypy
greedom has quit [Remote host closed the connection]
greedom has joined #pypy
Atque has joined #pypy
greedom has quit [Remote host closed the connection]