greedom has quit [Remote host closed the connection]
fotis has quit [Ping timeout: 260 seconds]
Julian has quit [Quit: leaving]
Julian has joined #pypy
Julian has quit [Quit: leaving]
fotis has joined #pypy
fotis has quit [Ping timeout: 260 seconds]
fotis has joined #pypy
fotis has quit [Ping timeout: 265 seconds]
lritter has joined #pypy
<mattip>
ctismer, ronan: we can call _type_realize() to reset the python-level type object from the c-PyTypeObject,
<mattip>
but I am not sure what will happen to the slots
<mattip>
are we supposed to recalculate all the inheritance of the object?
<mattip>
I assume not, since in the discussion above the mro did not change, only the type(obj)
<ctismer>
At least for my purposes, a minimum is needed. Feature switching replaces the type dict, only.
<mattip>
ahh, the entire __dict__ ? So that is all the slots __add__, etc
<mattip>
I guess tests are needed: what happens in CPython when the ob_type is switched
<mattip>
can you point me to the code that does pto->ob_type = &Shiboken.ObjectType
<mattip>
after PyType_Ready is called
<mattip>
I see one occurance in pyside-setup/sources/shiboken6/libshiboken/basewrapper.cpp in introduceWrapperType
<mattip>
nope, false positive
<ctismer>
I am currently re-implementing a lot, after Ronan‘s findings. Will take a few days. The PyType_FromSpec will be replaced for PyPy.
<ctismer>
The switching gave problems on Python 3.10 because of caching in getattr. PyType_Modified solved that very easily.
<mattip>
OK. I am willing to make a PyPy-specific change to PyType_Modified if needed
<ctismer>
Great! When is the next release planned? There are a few small omissions, too (will write that up, soon).
<mattip>
I just kicked off a set of builds that I hope will become rc1 as soon as it is ready
<mattip>
like by tomorrow
<ctismer>
But you can see them in the repo if you grep -i for `pyside-535‘`
<ctismer>
Ok, whow
fotis has joined #pypy
fotis has quit [Ping timeout: 252 seconds]
<ctismer>
If you grep for PYPY_VERSION you‘ll probably find more. The branch is not yet cleaned up for check-in. I am travelling today, not much possible ATM.
<mattip>
ok, thanks
fotis has joined #pypy
fotis has quit [Ping timeout: 252 seconds]
<mattip>
ctismer: I didn't find "pyside-535" in the pyside-setup repo at commit FETCH_HEAD from your directions,
<mattip>
but looking at the PYPY_VERSION ifdef conditionals, some of them are out of date
<mattip>
PyPy has PyDescr_NewMethod, has fixed the order of arguments to PyDescr_NewGetSet (in pypy3.7.4) but indeed is missing Py_GetProgramFullPath, Py_Initialize
<mattip>
If I "git grep -i pyside-535" I do see various comments in the code
<mattip>
anyhow, that's it for now, need to do other things
greedom has joined #pypy
Guest27 has joined #pypy
greedom has quit [Remote host closed the connection]
greedom has joined #pypy
<Guest27>
"PyPy ... is missing Py_GetProgramFullPath, Py_Initialize", shall wie open an issue ?
<mattip>
in principle all the functions in stubs.py (which includes these) are known to be missing
<mattip>
opening an issue will not do much unless we get more people than me to implement things
<Guest27>
this is the idea: expose the problem so more help may arrive
<mattip>
so perhaps open a tracking issue to implement all the functions in pypy/module/cpytext/stubs.py
fotis has joined #pypy
fotis has quit [Ping timeout: 260 seconds]
fotis has joined #pypy
fotis has quit [Ping timeout: 252 seconds]
greedom has quit [Remote host closed the connection]
fotis has joined #pypy
fotis has quit [Ping timeout: 252 seconds]
greedom has joined #pypy
slav0nic has joined #pypy
Atque has joined #pypy
<ctismer>
now in a hotel in Lübeck.
Dejan has joined #pypy
<ctismer>
mattip: you made great progress! The only missing non-trivial thing is Py_Initialize, maybe I can ignore this? I'd like to become a PyPy contributor, again - hopefully I can learn this, again 🙁
lritter has quit [Ping timeout: 252 seconds]
shlomif has joined #pypy
<shlomif>
Hi all! how should i profile mathsy pypy3 code ? «pypy3 -m vmprof -o out.log 764/euler_764_v1.py» says "unknown"s for most
<cfbolz>
shlomif: cprofile is a rough good first approximation
<cfbolz>
shlomif: I also played with scalene recently
<Guest2757>
the problem started apparently with the size of the binary moving to 29Mo. at first I thought it was just a failed nightly until mattip wrote is was nearing release today. so I checked
<LarstiQ>
does a non win64 build work?
<LarstiQ>
I guess I'll download a linux nightly
<LarstiQ>
Guest2757: I get the same problem, so seems a general issue
<LarstiQ>
with pypy-c-jit-103264-3b3e428f14b6-linux64 fwiw
<Guest2757>
the version that works was before the directory alignement to cPython standard (for site-packages and friends)
<cfbolz>
Guest2757: that fits your symptom
<cfbolz>
Of not finding a stdlib module
<cfbolz>
shlomif: I'll try to do some more digging later
<Guest2757>
lib-python of 44.9Mo went to Lib of 40.9Mo, a 4Mo loss
<shlomif>
cfbolz: thanks!
<Guest2757>
ensurepip shrinked from 5.6Mo to 2.2Mo
<shlomif>
cfbolz: profiling pypy3 code has been a long time wart here
<Guest2757>
ah, in previous version, there were 3 'pip' bundled, now only latest one pip-21.1.1'
<cfbolz>
shlomif: is cpython profiling really better?
<Guest2757>
latest one of previous working binary, as we are now at pip-21.2.4 or pip-21.1.3
<shlomif>
cfbolz: i don't think so
<Guest2757>
as a very ugly test, if I copy/past former 'lib-python' into today nightly... pypy3;exe starts again
* shlomif
is /away ; rest
<Guest2757>
and i get a prompt with "Python 3.8.12 (3b3e428f14b6, Sep 11 2021, 23:31:23)"
<cfbolz>
Guest2757: right
<Guest2757>
do I need to create a ticket ?
<cfbolz>
Guest2757: sounds useful yes
<Guest2757>
done
stkrdknmibalz has quit [Read error: Connection reset by peer]
Guest2757 has quit [Quit: Connection closed]
Guest27 has quit [Quit: Connection closed]
stkrdknmibalz has joined #pypy
greedom has quit [Remote host closed the connection]