2018-02-26 15:52
cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs:
https://botbot.me/freenode/pypy/ ) | use cffi for calling C | the secret reason for us trying to get PyPy users: to test the JIT well enough that we're somewhat confident about it
00:00
tbodt has joined #pypy
00:12
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
00:16
tbodt has joined #pypy
00:32
<
kenaan >
wlav cppyy-packaging 0df7710aad8b /pypy/module/_cppyy/: naming consistency w/ CPython/cppyy
00:32
<
kenaan >
wlav cppyy-packaging edb9132eda33 /pypy/module/_cppyy/test/: fix C++ warnings
00:32
<
kenaan >
wlav cppyy-packaging b74ad9bd1274 /: merge default into branch
00:33
<
kenaan >
wlav cppyy-packaging 60b72b97202f /: merge default into branch
00:33
<
kenaan >
wlav cppyy-packaging e98cf77f5b72 /pypy/module/_cppyy/: upgrade to backend 0.6.0
00:33
<
kenaan >
wlav cppyy-packaging 55cdebafb2de /pypy/module/_cppyy/interp_cppyy.py: translator fix
00:39
nimaje has quit [Remote host closed the connection]
00:39
nimaje has joined #pypy
00:58
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
00:59
jcea has quit [Quit: jcea]
01:04
tbodt has joined #pypy
01:25
ronan has quit [Ping timeout: 248 seconds]
01:33
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
01:41
tbodt has joined #pypy
01:59
_habnabit has joined #pypy
02:04
fryguybo1 has joined #pypy
02:07
fryguybob has quit [Ping timeout: 268 seconds]
03:38
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
03:40
tbodt has joined #pypy
03:44
tbodt has quit [Ping timeout: 248 seconds]
03:47
dddddd has quit [Read error: Connection reset by peer]
05:20
lazka has joined #pypy
05:44
_whitelogger has joined #pypy
06:00
fujisan3KFI0M has joined #pypy
06:00
<
fujisan3KFI0M >
_whitelogger lazka fryguybo1 _habnabit nimaje irclogs_io_bot norox zmt00 energizer pchiusano avakdh idnar TheAdversary Hasimir inhahe_ pepesza- demonimin mwhudson commandoline gclawes dustinm cjwelborn jacob22_ ulope tazle nanonyme Ganwell SunDwarf lapinot michaelgreene [Arfrever] redj ctismer Graypup_ abvi[m] yuvipanda bendlas dash agates pulkitg Civil [0__0] alcarithemad ronny mjacob_ infinite cadr_ danieljabailey mvantellingen bbot2 dan- glyph cerid
06:00
fujisan3KFI0M has quit [Client Quit]
06:10
oberstet has joined #pypy
07:31
lazka has quit [Quit: Leaving]
07:41
marr has joined #pypy
07:56
inhahe_ has quit [Ping timeout: 264 seconds]
08:00
asmeurer_ has joined #pypy
08:03
kenaan has quit [Read error: No route to host]
08:19
ronan has joined #pypy
08:24
energizer has quit [Quit: Leaving]
08:26
lazka has joined #pypy
08:36
asmeurer_ has quit [Quit: asmeurer_]
08:46
ronan has quit [Ping timeout: 248 seconds]
09:15
inhahe_ has joined #pypy
09:18
inhahe has quit [Ping timeout: 263 seconds]
09:24
lazka has quit [Quit: Leaving]
09:29
ronan has joined #pypy
11:27
jacob22__ has joined #pypy
11:28
jacob22_ has quit [Ping timeout: 256 seconds]
11:33
norox has quit [Quit: I'm quitting]
11:34
norox has joined #pypy
11:37
norox has quit [Remote host closed the connection]
11:37
norox has joined #pypy
12:16
dddddd_ has joined #pypy
12:17
dddddd_ is now known as dddddd
12:21
mcyprian has joined #pypy
12:31
inhahe_ has quit [Ping timeout: 260 seconds]
13:38
fryguybob has joined #pypy
13:41
fryguybo1 has quit [Ping timeout: 248 seconds]
13:54
lesshaste has joined #pypy
14:20
kenaan has joined #pypy
14:20
<
kenaan >
mwjackson py3.6 5781b52fc50f /: (ronan, mwjackson) added kwargs-only for sorted()
14:37
jcea has joined #pypy
14:41
lazka has joined #pypy
15:02
BotoX has joined #pypy
15:02
<
BotoX >
Hi, I want to switch from CPython + uWSGI + gevent to PyPy3
15:03
<
BotoX >
Installed uwsgi and gevent with pip, however on pypy uwsgi doesn't detect gevent
15:05
<
simpson >
BotoX: uwsgi can be a little hateful and picky, or at least that was the case a few years ago. Since you're just prototyping for now, maybe try PyPy with a simpler WSGI server first? But TBH I don't know which simple WSGI servers work on Python 3, so I'm really just assigning you more research, sorry. :T
15:06
<
BotoX >
This is on a production website with thousands of requests per second, I was hoping to lower the server load with PyPy
15:06
<
BotoX >
It works with gunicorn and gevent
15:07
<
BotoX >
I should probably try building uwsgi myself
15:08
<
simpson >
Sure. You might find that PyPy and a pure-Python WSGI server can perform as well as your production uwsgi. I dunno, just a possible way to move forward.
15:32
<
kenaan >
mwjackson py3.6 86478b4c1d71 /lib-python/3/test/test_platform.py: (ronan mwjackson) fixed platform tests that utilise sys._git we are not reverting the entire migration to _git i...
15:43
<
BotoX >
so the issue is that when installing gevent with pypy it only builds the pypy plugin
15:43
<
BotoX >
so I don't have the gevent plugin available at all
15:43
<
BotoX >
if I try to build uwsgi with pypy and gevent I get loads of undefined references like: _PyPy_Dealloc
15:44
<
BotoX >
I am using "Squeaky's portable Linux binaries.", do those come with the required libs for linking?
16:00
<
BotoX >
managed to build it, but the uwsgi plugin crashes haha
16:00
<
BotoX >
screw uwsgi I guess
16:00
<
BotoX >
*the gevent plugin crashes
16:01
<
tos9 >
BotoX: there's no real advantage to using uWSGI, yeah.
16:02
<
tos9 >
If you do use it though, make sure you're using the PyPy driver, though that's now somewhat old.
16:02
<
BotoX >
uWSGI seems to perform better under load than gunicorn according to benchmarks
16:02
<
tos9 >
BotoX: I've run gigantic prod deployments on PyPy+gunicorn for quite a few years.
16:02
<
tos9 >
BotoX: Those benchmarks never use PyPy
16:03
<
BotoX >
yeah yeah that's a given
16:03
<
BotoX >
uwsgi + pypy3 works fine
16:03
<
BotoX >
but not with gevent
16:03
<
BotoX >
it doesn't even build gevent by default
16:04
<
BotoX >
I successfully built it with gevent now and the gevent plugin segfaults
16:04
<
BotoX >
not going to sink more time into it, there are 400 issues open on uwsgi's github
16:04
<
BotoX >
and 65 pull requests too...
16:50
lazka has quit [Quit: Leaving]
16:51
mcyprian has quit [Quit: Leaving.]
17:00
asmeurer has joined #pypy
17:01
energizer has joined #pypy
17:05
asmeurer has quit [Ping timeout: 256 seconds]
17:12
illume has joined #pypy
17:19
lritter has joined #pypy
18:19
tbodt has joined #pypy
18:21
illume has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
18:47
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
18:55
asmeurer has joined #pypy
19:08
tbodt has joined #pypy
19:11
lritter has quit [Remote host closed the connection]
19:14
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
19:17
tbodt has joined #pypy
19:19
illume has joined #pypy
19:52
kipras`away is now known as kipras
20:02
illume has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
20:05
lazka has joined #pypy
20:07
<
lazka >
who is right, pypy or cpython? [][::-1] = [42]
20:08
<
lazka >
I'd say pypy
20:33
mattip has joined #pypy
20:40
lazka has quit [Quit: Leaving]
20:48
<
mattip >
NumPy merged PR #9998, so there should not be any more refcounting issues with nditer or anywhere else inside NumPy
20:50
<
ronan >
yay! congrats!
20:50
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
20:52
tbodt has joined #pypy
20:59
<
mattip >
ronan: still sprinting?
21:00
<
ronan >
mattip: yes, though I'm alone now
21:01
<
ronan >
(and waiiiitiiing for pandas to compile)
21:01
<
mattip >
thanks for the fix btw
21:01
<
ronan >
it's feels like just a hack though
21:02
<
mattip >
yes, not clear why the attach is not working correctly
21:02
<
ronan >
it's because the classes are pure-Python
21:03
<
ronan >
I'm not sure hanging a typedescr off a user class makes sense at all
21:04
<
mattip >
also, there is a warning about "builtins.type size changed", during importlib/_bootstrap.py
21:05
<
ronan >
I I thought that was the thing you fixed in Cython
21:08
jaffachief has joined #pypy
21:18
jaffachief has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
21:18
jaffachief has joined #pypy
21:21
jaffachief has quit [Client Quit]
21:22
jaffachief has joined #pypy
21:23
jaffachief has quit [Client Quit]
21:36
<
ronan >
looks like there's an issue with gc-hooks - the assert added in df2490d5d814 fires, causing a segfault in pandas
21:41
<
mattip >
it looks like the same thing, but is different. Trying to figure it out now
22:00
pf_moore has joined #pypy
22:00
<
ronan >
hmm, since a88536ae8554, it's no longer OK to fire the same action more than once
22:02
<
ronan >
which is quite inconvenient, given that actions must be prebuilt constants
22:10
AndrewBC has joined #pypy
22:18
<
ronan >
mattip: just opened #2805. I think it's a blocker for the release :(
22:51
<
mattip >
ronan: in the test you added to test_datetime, there is no cts imported. How does the test find it
22:52
<
ronan >
mattip: it comes from cdatetime
23:05
oberstet has quit [Ping timeout: 256 seconds]
23:06
marr has quit [Ping timeout: 276 seconds]
23:08
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
23:10
tbodt has joined #pypy
23:16
<
ronan >
pff, cython does something weird with array.array
23:17
<
ronan >
they define their own arrayobject in a way that's compatible with CPython's private definition
23:20
WhatisRT_ has joined #pypy
23:22
inhahe__ has joined #pypy
23:22
<
WhatisRT_ >
hi! I have a <cdata 'void *' ...>. How can I extract the address it is pointing to?
23:28
<
WhatisRT_ >
I have been trying for over and hour now and the only thing that works is converting it to a string, parsing the hex digits and converting that to an int
23:33
<
Alex_Gaynor >
WhatisRT_: `ffi.cast("uintptr_t", c_data)` should work I think
23:35
<
WhatisRT_ >
Alex_Gaynor: but the result of that is still a cdata
23:35
<
Alex_Gaynor >
WhatisRT_: err, then call `int()` on it
23:37
<
WhatisRT_ >
Alex_Gaynor: hm, that still didn't work: it doesn't give me the same number
23:48
<
WhatisRT_ >
when I print my cdata, I get <cdata 'void *' handle to <cdata 'void(*)(void *, uint8_t *, size_t, uint8_t * *, size_t *)' 0x101feb080>>
23:48
<
WhatisRT_ >
but the result of casting and converting to an int is 0x10397f818
23:49
<
WhatisRT_ >
I'm also really confused why these numbers have 9 digits
23:55
<
idnar >
leading zeroes probably aren't printed
23:55
<
idnar >
not sure about the rest though
23:56
<
WhatisRT_ >
yeah, I confused myself, for some reason I thought longs had 8 digits