<cfbolz>
amazigh: could you explain what exactly you are running so we can try to reproduce it?
<cfbolz>
mattip: I would commit a fix to #3644 to py3.7, is that ok?
catern has joined #pypy
<amazigh>
oops! sorry for the noise the problem was on my side, with sourcehut build service in the mix, instead of `poetry install` the following works: `poetry install --no-ansi --quiet`
<amazigh>
locally (!), it works.
<cfbolz>
cool
<cfbolz>
and that's not noise, it's rubber duck debugging ;-)
<LarstiQ>
conveniently not needing to explain what poetry and sourcehut are ;0
Guest96_ has joined #pypy
<mattip>
cfbolz: feel free to commit to the py3.* branches, I will cherry-pick/merge as needed to the releases
<cfbolz>
ok
<cfbolz>
mattip: also found another related 3.9 bug
<mattip>
it seems like corner cases are also most of the current test failures
<mattip>
maybe we should make a "pytest.mark.cornercase" decorator, and every now and then try to fix ten or twenty of them
<cfbolz>
:-)
<LarstiQ>
:)
Dejan has joined #pypy
<Cheery>
Hi
<Cheery>
considering to give constraints for the ukanren environment. It's a documented mod over existing implementation.
ebarrett has quit [Ping timeout: 260 seconds]
lritter has joined #pypy
slav0nic has quit [Remote host closed the connection]
greedom has joined #pypy
Dejan has quit [Quit: Leaving]
xcm has joined #pypy
xcm_ has quit [Read error: Connection reset by peer]
slav0nic has joined #pypy
otisolsen70_ has joined #pypy
otisolsen70 has quit [Ping timeout: 256 seconds]
greedom has quit [Remote host closed the connection]
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
otisolsen70_ has quit [Quit: Leaving]
<cfbolz>
ouch, just found a speed regression in the interpreter (the jit is fine) that I myself introduced at some point, around method calls to builtin types :-(
<Cheery>
I just realised I'm in a turing tarpit with this thing. It feels nasty to actually code with these combinators.
<cfbolz>
Cheery: hah :-)
<Cheery>
fst : (($0 ⊸ ⊥) + ($1 ⊸ ⊥) ⊸ ⊥) ⊸ $0
<Cheery>
then you wrangle a similar structure to the other side, you get an option which is an inverse of case statement.
<Cheery>
but they're too complicated for my brains to handle. I need to come up with an alternative.
otisolsen70 has joined #pypy
exarkun has joined #pypy
<exarkun>
Does this look familiar to anyone? AttributeError: cdata 'PROCESS_INFORMATION *' has no field 'dwFlags'
<exarkun>
mattip: That certainly explains the AttributeError. :)
<exarkun>
The code that tries to read it is in the PyPy stdlib though
<exarkun>
I am trying out 3.7 now
<mattip>
right, it is being passed in to CreateProcess
<mattip>
maybe from virtualenv\seed\wheels\acquire.py ?
Corbin has quit [Ping timeout: 240 seconds]
<mattip>
weird. The call to Popen in virtualenv does not use the startupinfo arg
<exarkun>
A PyPy 3.7 (7.3.7) run completed successfully
<mattip>
I think the error is refering to the left=hand side "si" not the startup_info, since the latter is a python object not a cffi object
<mattip>
but the error makes no sense, since si = _ffi.new("STARTUPINFO *") is two lines above
<mattip>
exarkun: could you try to add some debug printing around the error: is it si or startup_info that is in error?
<mattip>
"intermittent" suggests either the offending code is not called every time (i.e. the wheel is cached and does not need to be downloaded),
<exarkun>
It happens on CI and there is no caching in my CI configuration, so I /suppose/ whatever it downloads, it downloads every time
<mattip>
ahh, that makes it harder to debug
<exarkun>
And related to that, sadly, I'm not sure how much deeper testing I can do. I can't reproduce it locally, only on CI
<exarkun>
Strictly, I guess it happens in the step *after* Python gets installed in the environment, so there could be a step that rewrites part of the library with an instrumented version
<exarkun>
I definitely won't get to that today though
<exarkun>
If I see it again on PyPy 3.7 then I can probably justify spending some time getting more info
<mattip>
right. FWIW, we have dropped support for 3.6
<exarkun>
more reason not to investigate if it is 3.6-only I guess :)
<mattip>
is it a public CI run?
<exarkun>
Related to that, this is the PyPy that `actions/setup-python@v2` (GitHub Actions) gave me when I asked for "pypy3"