arigato 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 | mac OS and Fedora are not Windows
amz3 has joined #pypy
_whitelogger has joined #pypy
amz3 has quit [Ping timeout: 268 seconds]
dddddd has quit [Ping timeout: 268 seconds]
oberstet has joined #pypy
dddddd has joined #pypy
jcea has quit [Remote host closed the connection]
speeder39_ has joined #pypy
[Arfrever] has joined #pypy
stillinbeta has joined #pypy
<stillinbeta> Hello all! I'm dealing with a weird cython / pypy issue, and I'm wondering if anyone can help: https://github.com/marcusva/py-sdl2/issues/132
danieljabailey has joined #pypy
<mattip> stillinbeta: commented in the PR https://github.com/marcusva/py-sdl2/pull/133, you should be using a nightly
<mattip> the 6.0 releases are showing their age
amz3 has joined #pypy
speeder39_ has quit [Quit: Connection closed for inactivity]
lritter has joined #pypy
dddddd has quit [Ping timeout: 244 seconds]
<cfbolz> mattip: Re arm, with some luck David will be at the sprint (he set up all the arm builders originally)
<mattip> cfbolz: cool, will be nice to meet him
<cfbolz> mattip: arm will be a topic in any case, because Armin and maciek have arm64 stuff to do
stillinbeta has quit [Quit: Connection closed for inactivity]
<mattip> utf8-unicode-py3 fails to properly handle lower-case-final-sigma
<mattip> lower-sigma in the middle of a word is u'\u03c3', but at the end is u'\u03c2'
<mattip> we typically do some for ch in rutf8.Utf8StringIterator(value): <handle ch>; builder.append(ch)
<mattip> then at the end builder.build()
<mattip> so we need to either keep track of the last iteration of the loop and correct before appending,
<mattip> or teach the builder to check when it builds
<zaytsev> mattip: hey matti you won't believe be but i think i've found the problem
<mattip> zaytsev: hi
<zaytsev> mattip: i hope you speak fluent chinese - https://teratail.com/questions/80487
<mattip> "TypeScript was not added to PATH since a valid installation was not found" ?
<zaytsev> mattip: the thing is the damn microsoft installer doesn't set correct registry keys pointing to vcvarsall.bat. it also doesn't set the environment variable. it fact, this piece of shit doesn't get ANYTHING right. the file itself is there. your detection is based on variable. after this was set to correct value it worked.
<zaytsev> mattip: distutils want registry key, and this i didn't create and installer didn't either. so with a little help of my anonymous chinese friend i added it, and distutils seem to be able to find it now.
<zaytsev> mattip: oh it seems i mixed up chinese and japanese, whatever ;-) i meant this: レジストリ HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Microsoft\VisualStudio\14.0\Setup\VC の ProductDir に、
<zaytsev> 現在使っている Visual Studio の VC フォルダへのパス C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build を書き込む。
<zaytsev> mattip: yeah this hell is better be documented :( also now it's clear what was happening before
<zaytsev> mattip: i had 2017 thing installed alongside, because you didn't tell me to uncheck it. in itself, it's not a problem, but the installer manages to create correct registry keys for it. so pypy was picking up 14.0 and distutils 14.1, which didn't work
<zaytsev> mattip: anyways, i need to drive to $dayjob now :(
<zaytsev> mattip: the next 3.5 build should succeed. i'm all anticipation
<mattip> strange, on the other buildbot under
<mattip> HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\VisualStudio\14.0\Setup
<mattip> there is no VC key, and under
<mattip> HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\VisualStudio\14.0\VC
<mattip> there is no ProductDir key
<zaytsev> maybe you have it in HKLU, or search for ProductDir
<zaytsev> i ran find_vcvarsall(14.0) before, and it didn't find anything, then created the key and it worked
<zaytsev> mattip: another explanation that i found is that you need to install desktop development for c++ workload, but i didn't want to do that
<mattip> ok, well, whatever works. It is deep magic I guess
_whitelogger has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
Zaab1t has joined #pypy
Garen has quit [Read error: Connection reset by peer]
Garen has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<njs> hmm, this is interesting. apparently coverage.py gives nonsensical results on the pypy py36 branch
<njs> the comment on line 46 is marked as "uncovered", which makes no sense. And if you toggle the flags at the top, you can see that this is coming fromthe pypy_nightly_py3_6 build
<njs> (cc nedbat -- I don't know that it's worth your time to worrying about an unreleased pypy nightly build, but I figure you won't mind the heads-up)
<cfbolz> njs: ah, that's likely a bug in the wordcode implementation
Ai9zO5AP has quit [Ping timeout: 244 seconds]
<cfbolz> thanks for the heads up
Ai9zO5AP has joined #pypy
<njs> sounds plausible
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<mattip> njs: is someone using py3.6, and it works?
<njs> mattip: trio has a CI builder that runs on the py3.6 branch nightly
<njs> mattip: it's worked fine for ages
<mattip> nice. Does that have a typical web-based python usage profile: templating, sockets ... ?
<mattip> i wonder how complete py3.6 is
<cfbolz> we all wonder
<cfbolz> the answer is a resounding "quite complete"
antocuni has joined #pypy
<njs> mattip: trio is heavily biased towards low-level I/O stuff. No templating, lots and lots of sockets/ssl/weird signal hacks/etc.
<cfbolz> still impressive that it works at all on pypy ;-)
<njs> (the usual example I give is: the tests for trio's ssl support haven't found any bugs in trio recently, but they have found bugs in cpython's ssl module, pypy's ssl module, pyopenssl, openssl, and arguably the TLS 1.3 spec.)
<mattip> njs: :)
<cfbolz> haha
<cfbolz> njs: you should tweet that
<fijal> njs: yay!
<njs> I should probably write a blog post about that TLS 1.3 thing, it is exactly the kind of arcane and barely-relevant content that my audience demands
<njs> (of course by "demands" I mean "absolutely no-one demands but I like writing about anyway")
<fijal> njs: all of that can go viral on twitter btw ;-)
<fijal> for what is worth, I demand arcane and barely relevant content from you!
<njs> heh
<njs> well, you can get the preview here: https://github.com/openssl/openssl/issues/7967
<njs> (maybe https://github.com/openssl/openssl/issues/7948 is necessary for context, idk)
<fijal> turns out networking is hard
<fijal> or more importantly, "why does apt not use https" and discussion that follows
<njs> fijal: I haven't seen that one, no
<fijal> so I'm too uninformed on that
<fijal> I can see both sides being right
<fijal> so would be cool if someone who knows about security will explain it to me :]
<fijal> this is the point of contention
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<hruske> turns out that if your link is real bad, having TLS with all kinds of replay preventions also prevents caching
<hruske> thus making most online stuff unusable
<kenaan> mattip unicode-utf8-py3 0e658f7c342f /pypy/objspace/std/: calculate padding from unicode length, not utf8 length
<kenaan> mattip unicode-utf8-py3 1142ef536c49 /pypy/objspace/std/: test, fix capitalize with sigma in final position
<kenaan> mattip unicode-utf8-py3 d529e654cfb5 /pypy/: test, fix for suppressing multiple error output bytes
<njs> fijal: sounds legit to me
<nedbat> njs: let me know if you need help looking into it
<njs> nedbat: well, I just disabled coverage reporting for the pypy py3.6 nightlies for now
<nedbat> if only everyone did that when they found a coverage bug :)
<cfbolz> it's much more likely a pypy bug
<kenaan> mattip unicode-utf8-py3 3f8c26309cd2 /pypy/interpreter/unicodehelper.py: try not special-casing mac OS
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/5949 [mattip: force build, unicode-utf8-py3]
<bbot2> Started: http://buildbot.pypy.org/builders/own-linux-x86-64/builds/7223 [mattip: force build, unicode-utf8-py3]
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/4176 [mattip: force build, unicode-utf8-py3]
dddddd has joined #pypy
arigato has quit [Read error: Connection reset by peer]
bbot2 has quit [Read error: Connection reset by peer]
bbot2 has joined #pypy
<kenaan> cfbolz unicode-utf8 b6331207f8b9 /pypy/objspace/std/: change UnicodeDictStrategy to store wrapped unicode objects as keys. this lifts the restriction that they are...
<kenaan> cfbolz unicode-utf8 df16e768b3bb /pypy/: make json decoding efficient again
Zaab1t has quit [Ping timeout: 272 seconds]
antocuni has quit [Ping timeout: 268 seconds]
<kenaan> cfbolz unicode-utf8 4bd5dbda0076 /TODO: done by df16e768b3bb
Zaab1t has joined #pypy
marky1991 has joined #pypy
<mattip> we might want to upload latest builds to bitbucket
<mattip> if CI start downloading latest from buildbot, buildbot.pypy.org is going to get hit with alot of downloads
<cfbolz> mattip: quite a few ci systems download from buildbot already, I think
<mattip> fijal ^^^
jcea has joined #pypy
marky1991_2 has joined #pypy
<mattip> that;s weird, bbot2 said it started those jobs, but then disconnected and reconnected and now they are gone
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/5949 [mattip: force build, unicode-utf8-py3]
<bbot2> Started: http://buildbot.pypy.org/builders/own-linux-x86-64/builds/7223 [mattip: force build, unicode-utf8-py3]
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/4176 [mattip: force build, unicode-utf8-py3]
marky1991 has quit [Ping timeout: 246 seconds]
<mattip> even the build numbers did not increment
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<cfbolz> mattip: maciek told me he had to restart the build master because it ran out of file descriptors
marky1991_2 has quit [Ping timeout: 268 seconds]
<mattip> ahh, thanks
marky1991 has joined #pypy
marky1991 has quit [Remote host closed the connection]
marky1991 has joined #pypy
adamholmberg has joined #pypy
antocuni has joined #pypy
marky1991 has quit [Remote host closed the connection]
marky1991 has joined #pypy
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/5949 [mattip: force build, unicode-utf8-py3]
xcm has quit [Remote host closed the connection]
Rhy0lite has joined #pypy
xcm has joined #pypy
<fijal> mattip: hey matti. I had to reboot the buildmaster machine
<fijal> sorry for inconvinience but it run out of ptys
<fijal> not sure why and how
<fijal> but seems critical :)
antocuni has quit [Ping timeout: 250 seconds]
adamholm_ has joined #pypy
adamholmberg has quit [Ping timeout: 250 seconds]
<mattip> ok, cool. Good thing you noticed
marky1991 has quit [Remote host closed the connection]
<fijal> well I could not ssh
<mattip> maybe someone uploading a large vmprof?
<fijal> hard to say
<bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/4176 [mattip: force build, unicode-utf8-py3]
<fijal> vmprof.com should really go somewhere else
fryguybo1 has quit [Quit: leaving]
fryguybob has joined #pypy
<bbot2> Failure: http://buildbot.pypy.org/builders/own-linux-x86-64/builds/7223 [mattip: force build, unicode-utf8-py3]
<mattip> I don't understand why cpython isn't consistent with sigma https://en.wikipedia.org/wiki/Sigma
<mattip> on the one hand, u'A\u0345\u03a3'.swapcase() == u'a\u0399\u03c2', which I understand, that is the GREEK SMALL LETTER FINAL SIGMA
<mattip> but what is going on with u'\u0345\u03a3'.swapcase() == u'\u0399\u03c3', which is the GREEK SMALL LETTER SIGMA
<mattip> is u0345 special?
<mattip> ahh, indeed https://codepoints.net/U+0345 says it is COMBINING GREEK YPOGEGRAMMENI
<mattip> and so I guess shouldn't be counted towards converting 03a3 to 03a2 at the end of a word
<mattip> s/converting/determining if to use/
<cfbolz> Snow snow!
marky1991 has joined #pypy
marky1991 has quit [Ping timeout: 245 seconds]
jacob22__ has quit [Quit: Konversation terminated!]
oberstet has quit [Remote host closed the connection]
Ai9zO5AP has quit [Ping timeout: 246 seconds]
Ai9zO5AP has joined #pypy
<mattip> cpython has special casing for sigma in lower_ucs4 and a dedicated handle_capital_sigma to decide 03a3 or 03a2
<mattip> tumbleweed: are there arm64 pypy packages on debian?
<ronan> mattip: the sigma issue is rather stupid, but it's part of the Unicode standard, so everybody needs to implement a special case for it
<mattip> tumbleweed: see the last line of this issue https://bitbucket.org/pypy/pypy/issues/2852
<mattip> ronan: like most unicode things it is clear now, but I will forget why I wrote the code in another month
<mattip> Arabic and Hebrew also have final letters, but as far as I know they don't have upper-lower cases, so it is easier to reason about
<kenaan> rlamy default eec4eefc8618 /rpython/jit/metainterp/test/test_zvector.py: Reduce indirection in generating hypothesis tests
<ronan> mattip: AFAIK, the handling of final sigma is a legacy issue. For every other case, they decided to treat it as glyph variants.
<kenaan> rlamy default 33fe32bc9404 /rpython/jit/metainterp/test/test_zvector.py: Delete confusing unused method
kipras has joined #pypy
lritter has quit [Ping timeout: 246 seconds]
marky1991 has joined #pypy
<tumbleweed> mattip: there are
<ronan> I'm trying to debug the test_site failure on py3.5, and something weird is going on: we have modules that come from nowhere in sys.modules, with paths apparently set up at translation-time
<ronan> I'm surprised nothing broke
<cfbolz> nedbat: I'm running the coverage.py tests. fwiw, both pypy 3.5 and 3.6 have quite a few failures, but 3.6 in particularly fails test_arcs.py a lot (whereas 3.5 passes it)
<nedbat> cfbolz: do you have a link to the failures?
<cfbolz> nedbat: no, just ran it locally. should I paste the names of the failing tests?
<nedbat> cfbolz: i'd be curious to see the full failures.
<mattip> tumbleweed: were there problems translating it?
<cfbolz> nedbat: including tracebacks? can do
<nedbat> cfbolz: it's not unusual to have to adjust the tests based on changes in the python implementation
<cfbolz> nedbat: do you run the tests with some version of pypy?
<nedbat> cfbolz: locally i run with: PyPy 3.5.3 (pypy 6.0.0.final.0)
<cfbolz> and that passes?
<nedbat> cfbolz: yes
<mattip> tumbleweed: I ask because we are thinking of how best to build for ARM, ARM64. We currently use this setup
<mattip> tumbleweed: and are wondering if there is a simpler/better way to do it
<mattip> cfbolz: I think njs was running nightly py3.6
<cfbolz> mattip: yes, I am running both to see the differences
<tumbleweed> mattip: we do only native translation
<mattip> tumbleweed: so you can do ARM32? I thought translation takes more than 4GB
<tumbleweed> yep, we do
<tumbleweed> want me to measure the usage?
<mattip> sure. also how long does it take?
<tumbleweed> native translation is slow, of course http://buildd.debian.org/pypy
<tumbleweed> 10 hours for translation + tests on armhf
<cfbolz> nedbat: is it ok to just run pytest? or is that bound to run into trouble and I should hook up pypy3.6 into tox?
<mattip> tumbleweed: cool. Any idea what hardware is behind that build slave?
<tumbleweed> our hardware is listed on https://db.debian.org/machines.cgi
<tumbleweed> looks like they are all Marvell Armada 370/XP CPU @ 1.6GHz on a Marvell MV78460 SoC Development Board (ARM v7)
<mattip> +1
<cfbolz> mattip: translation fits into 4gb on 32bit
<cfbolz> nedbat: sorry, that was a silly question, I'll just do the right thing and go via tox
Rhy0lite has quit [Quit: Leaving]
<mattip> right, but most arm boards only come with 2GB, it seems these are beefier (and probably more expensive)
<cfbolz> yeah
agronholm has quit [Ping timeout: 250 seconds]
<cfbolz> nedbat: here are the results (3.5 first, then 3.6) https://www.irccloud.com/pastebin/VQ01d8a1/
<nedbat> cfbolz: it looks like 1) there's something changed about non-ascii in 3.5, and 2) the line number for compiled versions of multi-line constructs changed in 3.6
<cfbolz> nedbat: yes, it's definitely a pypy 3.6 bug, it behaves completely wrong in pdb too
marky1991 has quit [Remote host closed the connection]
marky1991 has joined #pypy
agronholm has joined #pypy
<mattip> it seems I actually added code to handle sigma properly ~6 months ago, and forgot all about it
<cfbolz> :-(
<cfbolz> mattip: but only on py3.5?
Zaab1t has quit [Quit: bye bye friends]
<cfbolz> Waaaat
<cfbolz> I just got asked to review a paper on 'Nurses just-in-time training'
senyai has quit [Ping timeout: 240 seconds]
marky1991 has quit [Ping timeout: 250 seconds]
jcea has quit [Remote host closed the connection]
jcea has joined #pypy
marky1991 has joined #pypy
<simpson> Nice. Is it any good?
marky1991 has quit [Remote host closed the connection]
marky1991 has joined #pypy
moei has quit [Quit: Leaving...]
<tos9> "Patient is in cardiac arrest"
<tos9> <Hands nurse textbook>
<simpson> The idea is that, if a nurse is generally qualified but doesn't have training in a specific technique which is specialized to a particular clinic or situation, then they can get that training on the job rather than in a classroom.
<cfbolz> Well, I declined the review
<cfbolz> So i couldn't look at the paper, only the abstract
<cfbolz> Clearly some pattern matching gone wrong to try to get me to review it 😜
<simpson> Compare and contrast with things like runbooks, maybe? I'm making a note to chat with some relatives in nursing about this.
<cfbolz> Heh
antocuni has joined #pypy
jacob22__ has joined #pypy
kipras has quit [Read error: Connection reset by peer]
adamholm_ has quit [Remote host closed the connection]
xcm has quit [Remote host closed the connection]
xcm has joined #pypy