cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end ) | use cffi for calling C | if a pep adds a mere 25-30 [C-API] functions or so, it's a drop in the ocean (cough) - Armin
<bbot2_> Started: http://buildbot.pypy.org/builders/own-linux-s390x/builds/1620 [mattip: Forced by command line script, release-pypy3.7-v7.x]
<bbot2_> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-s390x/builds/1454 [mattip: Forced by command line script, release-pypy3.7-v7.x]
todda7 has joined #pypy
lazka has quit [Quit: Ping timeout (120 seconds)]
lazka has joined #pypy
jcea has quit [Ping timeout: 245 seconds]
dustinm has quit [Quit: Leaving]
dustinm has joined #pypy
<bbot2_> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-linux-s390x/builds/1454 [mattip: Forced by command line script, release-pypy3.7-v7.x]
Gustavo6046 has joined #pypy
arigo_ has quit [Quit: ZNC - http://znc.in]
arigo has joined #pypy
arigo is now known as arigato
michelp has quit [Ping timeout: 245 seconds]
cfbolz has quit [Read error: Connection reset by peer]
EWDurbin has quit [Read error: Connection reset by peer]
cfbolz has joined #pypy
michelp has joined #pypy
EWDurbin has joined #pypy
<bbot2_> Failure: http://buildbot.pypy.org/builders/own-linux-s390x/builds/1620 [mattip: Forced by command line script, release-pypy3.7-v7.x]
lritter has joined #pypy
oberstet_ has quit [Quit: Leaving]
todda7 has quit [Ping timeout: 240 seconds]
<cfbolz> we really need to look at bigger applications more :-(
<cfbolz> I am finding again a million papercuts by studying the font stuff
<mattip> in some sense that is a good sign of moving from a research project to production
<cfbolz> yeah
<cfbolz> about 30% is C execution, that part is unclear we can win on
<bbot2_> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/159 [mattip: force build, win64-cpyext-default]
<bbot2_> Started: http://buildbot.pypy.org/builders/own-win-x86-64/builds/235 [mattip: force build, win64-cpyext-default]
<mattip> would hpy help?
<cfbolz> mattip: I don't know, I am just seeing it in the profile
<cfbolz> heh, bit sad: pypy takes 12s on the C++ code, cpy 5s. the total runtime is the same (~30s), meaning that the JIT helps a lot, but the gains are offset by the cpyext slowdowns
<cfbolz> in other news, I should really actually finish some of my branches ;-)
<mattip> also other news: if HPy finds someone to do a logo/icon, PyPy could use a new ico file in pypy/goal
<mattip> the one I made from the logo is not that good
<cfbolz> that stuff's hard indeed :-(
<nulano> mattip, I'm not sure it's worth fixing cpyext on PyPy2 for win64
<nulano> there are a lot more issues than on py3
<mattip> nulano: I doubt anyone is actually using it in practice, but not having it messes up CI runs
<mattip> let's see if the branch translates
<antocuni> cfbolz: what you find matches my experience. Our C performance is so awful that eats everything that we win in pure-python. This was of the reasons why I wanted to start hpy :)
<nulano> mattip, at the very least I wouldn't release it without carefully checking the API. For example, py2 has 32-bit hashes even on win64 (unlike py3)
<mattip> ahh
<cfbolz> antocuni: I am less pessimistic in practice
<mattip> ok, I will ask the cibuildwheel people if they can deal with missing cpyext on win64-python2.7
<antocuni> cfbolz: what do you mean?
<mattip> nulano ^^^
<cfbolz> antocuni: there's a lot of pure python code around
<cfbolz> but yes, we need to break even on C code
<cfbolz> in the long run
<antocuni> yes, that's my point
<cfbolz> but still, we need to fix the papercuts too
nimaje has quit [Quit: WeeChat 3.0.1]
<antocuni> sure, we need to to both
nimaje has joined #pypy
<nulano> mattip, I wonder whether effectively removing cpyext from py2 on win might surprise someone given the faq states that pypy2 "will be around 'forever'"
<nulano> but that can be justified by the fact that rpython doesn't need cpyext
<mattip> nulano: we can document it. I really doubt anyone will call us out on it, and if they want it they can sponsor the work
<nulano> sounds good
<cfbolz> sounds good to me as well
<cfbolz> antocuni: this is the corresponding cython code that is slow in this case, if you're interested: https://github.com/googlefonts/compreffor/blob/main/src/cython/_compreffor.pyx
<antocuni> cfbolz: it looks like a perfect task for cffi
<cfbolz> that too ;-)
<nulano> mattip, how's this icon? https://pasteboard.co/JUnTTwA.png
<nulano> haven't tried translating yet, but I would be surprised if the exe looks much worse than that
<cfbolz> much better definitely!
<mattip> nulano: looks much better
<nimaje> (why does a site to simply show an image need js (and that much at that)?)
tsaka__ has joined #pypy
<nulano> new icon in merge request 811
<nulano> fun fact: after uploading an image to heptapod fails, the try again button instead saves changes and closes the MR editor
<cfbolz> nice, the gitlab diff for images is pretty slick
<cfbolz> I clicked merge ;-)
<mattip> openssl has a bugfix release, so I guess we should update all the builds to openssh-1.1.1k
<nulano> win currently uses the build from https://github.com/python/cpython-bin-deps/tree/openssl-bin, which is still openssh-1.1.1g
<nulano> actually, there's a tag for openssh-1.1.1i, but not k yet
<mattip> it came out in response to some CVEs this week
<mattip> because we do portable builds, I need to regenerate the docker images and restart the buildslaves
<bbot2_> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/159 [mattip: force build, win64-cpyext-default]
YannickJadoul has joined #pypy
<mattip> nulano: even though I modified tp_hash to return a longlong, tests pass
<mattip> I guess the tests are consistent with themselves, or maybe we never really test tp_hash well
<bbot2_> Failure: http://buildbot.pypy.org/builders/own-win-x86-64/builds/235 [mattip: force build, win64-cpyext-default]
Gustavo6046 has quit [Ping timeout: 252 seconds]
Gustavo6046_ has joined #pypy
Gustavo6046_ is now known as Gustavo6046
<bbot2_> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/160 [nulano: rebuilt tcl/tk, py3.7]
<nulano> mattip, I think the tcl theme issue should be fixed
<nulano> I'm guessing the test for tp_hash (if there is one) assumes that it will be Signed, same as other platforms
dnshane has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
dnshane has joined #pypy
<bbot2_> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-64/builds/160 [nulano: rebuilt tcl/tk, py3.7]
astronavt has quit [Quit: ...]
astronavt has joined #pypy
YannickJadoul has quit [Quit: Leaving]
r1ch has joined #pypy
jcea has joined #pypy
lritter has quit [Ping timeout: 246 seconds]
<bbot2_> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5766 [nulano: rebuilt tcl/tk, py3.7]
tsaka__ has quit [Ping timeout: 240 seconds]
tsaka__ has joined #pypy
mgornyz has joined #pypy
mgorny has quit [Ping timeout: 250 seconds]
<bbot2_> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5766 [nulano: rebuilt tcl/tk, py3.7]
mgornyz has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
mgorny has joined #pypy
<cfbolz> mattip (logs): it would be cool if we could still cherry-pick 7cdc64becf9a for the py3.7 release branch
r1ch has quit [Quit: bye!]