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
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
Dejan has quit [Quit: Leaving]
adamholmberg has joined #pypy
adamholmberg has quit [Read error: Connection reset by peer]
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
tsaka__ has quit [Ping timeout: 240 seconds]
tsaka__ has joined #pypy
tsaka__ has quit [Ping timeout: 246 seconds]
jcea has quit [Quit: jcea]
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
BPL has quit [Quit: Leaving]
tsaka__ has joined #pypy
tsaka__ has quit [Ping timeout: 246 seconds]
epony has joined #pypy
<tumbleweed> I'm going to start building the docs that we ship in debian, with python 3 sphinx, so without autodoc and the other extensions (e.g. config). So that debian can drop the python 2 sphinx module
<tumbleweed> that stuff looks mostly useful for developers rather than users
<tumbleweed> I can't help wondering how much longer readthedocs is going to support 2.7, either
tsaka__ has joined #pypy
epony has quit [Quit: reconf]
epony has joined #pypy
epony has quit [Remote host closed the connection]
tsaka__ has quit [Ping timeout: 264 seconds]
epony has joined #pypy
jvesely has quit [Quit: jvesely]
_whitelogger has joined #pypy
_whitelogger has joined #pypy
dansan_ is now known as dansan
<mattip> tumbleweed: thanks for the update
<mattip> I guess autodoc won't work until you can import rpython/pypy since it needs to traverse parsed code
<mattip> so the patch seems legit
<mattip> dunno what we will do, maybe self-host the documentation :(
<mattip> YannickJadoul (for the logs): updating python2 on the buildbot solved building the cffi modules
<mattip> YannickJadoul (for the logs): there are still link errors, even though the buildbot is using the same g++ as you.
<mattip> can you find that missing library "ld: library not found for -lstdc++" and figure out how you have it when the buildbot doesn't?
dddddd has quit [Remote host closed the connection]
xcm has quit [Ping timeout: 264 seconds]
xcm has joined #pypy
Dejan has joined #pypy
Dejan has joined #pypy
Dejan has quit [Changing host]
dnshane has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
dnshane has joined #pypy
<nimaje> hi, I try to update the freebsd ports of pypy{,3} and for some reason when building pypy3 with pypy it can't find _curses_cffi_check, but there is a work/pypy3.6-v7.3.0-src/lib_pypy/_curses_cffi_check.pypy36-pp73-freebsd12.so (here is a build log https://0x0.st/iaGY.log) (works when building with cpython)
<mattip> Maybe the try/except block in _curses_build.py should be catching ModuleNotFoundError as well?
<mattip> nimaje: does ldd _curses_cffi_check.pypy36-pp73-freebsd12.so show everything is OK? Maybe the linker succeeds but the library is missing?
tsaka__ has joined #pypy
<nimaje> mattip: it seems fine https://0x0.st/iaFv.txt
<mattip> weird. On the other hand, I see that ldd pulls in both libcursesw and libpanelw, where the check is only for one of them at a time
<mattip> maybe the error is from a previous invocation of _curses_build and the file you see is from a second call?
<nimaje> both libcursesw and libpanelw is maybe because of /usr/lib/libpanelw.so.5: libncursesw.so.8 => /lib/libncursesw.so.8 (0x800673000)
* mattip off for a bit, bbl
<nimaje> could be that that _curses_cffi_check was from a cpython build
tsaka__ has quit [Ping timeout: 246 seconds]
YannickJadoul has joined #pypy
<YannickJadoul> mattip: Ah, great. One step forward. I'm not entirely sure where to check these link flags, but I'll have a look
dddddd has joined #pypy
<YannickJadoul> But now there seems to be a success nightly build? So where do I look for this link error?
lritter has joined #pypy
adamholmberg has joined #pypy
jcea has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
tsaka__ has joined #pypy
camelCaser has quit [Ping timeout: 250 seconds]
camelCaser has joined #pypy
jvesely has joined #pypy
<mattip> you can search for "linker command failed"
<mattip> nimaje: ahh, that is strange
<mattip> you can delete that so and try to run the script again "cd lib_pypy; pypy3 _curses_build.py"
<mattip> Given the push in conda to create pypy3.6 packages, I would like to push forward with 7.3.1 soon, without 3.7 alpha
<mattip> cfbolz, ronan: thoughts?
yajadoul has joined #pypy
<cfbolz> mattip: yes, 3.7 can wait for a bit
YannickJadoul has quit [Ping timeout: 250 seconds]
<nimaje> ok, now it built, but all builds were in a fresh and clean environment, so no idea why it works now or why it wasn't before
yajadoul has quit [Quit: Leaving]
YannickJadoul has joined #pypy
<YannickJadoul> mattip: Thanks! I'll have a look in a bit
<mattip> nimaje: ???
<Dejan> so.8 ?
<nimaje> mattip: I ran the build again with pypy to get the _curses_cffi_check.pypy36-pp73-freebsd12.so but now it succeded instead of failing
<Dejan> I thought ncurses SO versions follow the major release versions
<Dejan> at least they do on Fedora/CentOS/RedHat
<mattip> nimaje: yeah, that seems to happen on the buildbots too. There they run a translation which at the end builds the cffi packages,
<mattip> and then in a separate step call package, which again builds the cffi packages
<mattip> I seem to remember that error from soemwhere
<mattip> but since the second call succeeds, well, dunno what to do
<nimaje> well, the whole build was done again including translation starting from a fresh freebsd installation
<mattip> whohoo, conda forge blog post about the PyPy support https://conda-forge.org/blog//2020/03/10/pypy
jvesely has quit [Quit: jvesely]
<mattip> already retweeted, I know but I just saw it
<antocuni> mattip: hoorray for conda/pypy :)
<mattip> scipy still does not build, but numpy does
<antocuni> doesn't it build for conda-specific reason, or because it's just broken on pypy?
<mattip> it needs 7.3.1
<antocuni> btw, is there any chance that numpy will start to build its own pypy wheels?
<mattip> working on it, not simple. Two blockers so far: https://github.com/numpy/numpy/issues/15796
<mattip> in neither is it clear that PyPy is at fault (so far)
<antocuni> mattip: re #15775. I didn't look at details at all, but might it be caused by the fact that if we have a pypy->cpyext->pypy reference cycle, our GC is currently unable to collect the objects?
<mattip> is there a way to confirm/verify that such objects exist?
<antocuni> I recently wrote some hacky code to investigate in that direction, let me find it
<mattip> I tried adding all the python objects I could find to a weakref.ref list, but after enough gc.collects they all were dead
<nimaje> ok and now the build failed again, ldd work/pypy3.6-v7.3.0-src/lib_pypy/_curses_cffi_check.pypy36-pp73-freebsd12.so says https://0x0.st/iBrF.txt
camelCaser has quit [Read error: Connection reset by peer]
camelCaser has joined #pypy
<antocuni> this is a small cpyext extension which calls tp_traverse (it's basically a copy of CPython's gc_get_referents)
<antocuni> and a small snippet which calls tp_traverse on all the cpyext objects which are alive
<antocuni> this should let you to understand whether there is any numpy-related object which keeps alive "something"
<antocuni> note that it's not a full cycle detection system, because e.g. it doesn't consider gc->cpyext links
<antocuni> but it's probably possible to write the logic using gc.get_referents if you need
<mattip> thanks!
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
BPL has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
tsaka__ has quit [Quit: Konversation terminated!]
tsaka__ has joined #pypy
xcm has quit [Ping timeout: 264 seconds]
<cfbolz> hm, I am seeing a lot of warnings in cpython tests like this:
<cfbolz> threading_cleanup() failed to cleanup 0 threads (count: 0, dangling: 2)
<cfbolz> have they ever been there?
<cfbolz> is that a new thing since the GIL changes?
xcm has joined #pypy
xcm is now known as Guest18619
Guest18619 has quit [Killed (tepper.freenode.net (Nickname regained by services))]
xcm has joined #pypy
FullTiltProspect has joined #pypy
<YannickJadoul> mattip: This seems useful
<YannickJadoul> `xcode-select -p` results in `/Library/Developer/CommandLineTools` for me, so maybe `--install`ing the latest version might help? Or `xcode-select --reset`?
marky1991 has joined #pypy
<YannickJadoul> Ha, that took ages, but I found something useful, perhaps. This is the output of `gcc -v -x c++ /dev/null -fsyntax-only` (which is actually mac's adapted clang, and not GCC, but ...): https://pastebin.com/2ykyTV18
<mattip> cfbolz: I have not seen that before. Are there particular tests that trigger the warning?
FullTiltProspect has quit [Ping timeout: 256 seconds]
<mattip> YannickJadoul: mayve we need to add -stdlib=libc++ to the macOS Makefile
<mattip> sorry, the macOS cffi recompile build scripts
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5197 [mattip: force build, release-pypy3.6-v7.x]
dansan has quit [*.net *.split]
mattip has quit [*.net *.split]
glyph has quit [*.net *.split]
mgedmin has quit [*.net *.split]
tos9 has quit [*.net *.split]
rjarry has quit [*.net *.split]
glyph has joined #pypy
mgedmin has joined #pypy
tos9 has joined #pypy
dansan has joined #pypy
mattip has joined #pypy
<bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5197 [mattip: force build, release-pypy3.6-v7.x]
arigato has quit [Ping timeout: 240 seconds]
arigo has joined #pypy
marky1991 has quit [Remote host closed the connection]
rjarry has joined #pypy
jvesely has joined #pypy
marky1991 has joined #pypy
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5198 [mattip: force build, release-pypy2.7-v7.x]
Smigwell has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<cfbolz> mattip: eg test_asyncio, but on 3.7
<cfbolz> I suppose I should try 3.6
<YannickJadoul> mattip: Maybe, but normally, that should be the default, I'd say. Or at least it should not override platforms
<YannickJadoul> So maybe `export CPPFLAGS=-stdlib=libc+` on the buildbot, then?
<mattip> I wonder what in those tests needs c++
<bbot2> Exception: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5198 [mattip: force build, release-pypy2.7-v7.x]
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5199 [mattip: force build, release-pypy3.6-v7.x]
<Dejan> I did not try hypercorn on latest 3.7 build... I guess that would make some people happy.
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
whitewolf has joined #pypy
Dejan has quit [Quit: Leaving]
<bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/5199 [mattip: force build, release-pypy3.6-v7.x]
Taggnostr has quit [Remote host closed the connection]
Taggnostr has joined #pypy
lritter has quit [Quit: Leaving]
adamholmberg has quit [Remote host closed the connection]
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
XorSwap has joined #pypy
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
adamholmberg has joined #pypy
adamholmberg has quit [Remote host closed the connection]
YannickJadoul has quit [Quit: Leaving]
BPL has quit [Quit: Leaving]
adamholmberg has joined #pypy
adamholmberg has quit [Ping timeout: 264 seconds]
jvesely has quit [Read error: Connection reset by peer]
jvesely has joined #pypy
BPL has joined #pypy
XorSwap has quit [Remote host closed the connection]
XorSwap has joined #pypy