00:04
lritter has quit [Quit: Leaving]
01:35
_whitelogger has joined #pypy
02:02
jcea has quit [Quit: jcea]
02:41
Rhy0lite has quit [Quit: This computer has gone to sleep]
03:52
xcm has quit [Remote host closed the connection]
06:55
dustinm- has quit [Quit: Leaving]
06:56
dustinm has joined #pypy
07:42
oberstet has joined #pypy
07:55
<
mattip >
anyone want to comment on the fix-fstring-double-braces topic branch? LGTM, but there are so many ways to detect this ...
08:24
<
mattip >
arigo: in general, when we go from utf8 to unicode and back via s.decode('utf-8') or u.encode('utf-8') in RPython code
08:24
<
mattip >
how does it actually translate?
08:25
<
mattip >
if all we want is a wchar utf16 convervsion for win32 system calls, what is the fastest way to do that?
08:26
<
mattip >
there is rffi.scoped_utf82wcharp(source, unicode_len)
08:30
<
mattip >
which uses "for ch in rutf8.Utf8StringIterator(utf8): w[index] = unichr(ch)" underneath to convert
09:15
camelCaser has quit [Ping timeout: 265 seconds]
09:17
camelCaser has joined #pypy
10:00
Dejan has joined #pypy
10:02
Dejan has quit [Remote host closed the connection]
10:22
astrojl_matrix has quit [Ping timeout: 260 seconds]
10:23
altendky has quit [Ping timeout: 256 seconds]
10:24
astrojl_matrix has joined #pypy
10:24
altendky has joined #pypy
10:34
TheNewbie has joined #pypy
10:53
epony has quit [Quit: reconfigure-now]
10:53
epony has joined #pypy
11:16
<
arigo >
Mattip: unsure, but look at module/posix for example
11:17
<
mattip >
right, I understand the practice, I was more interested in the theory
11:17
<
mattip >
how does s.decode('utf-8') get translated
11:23
<
arigo >
see rpython/rstr.py or rpython/lltypesystem/rstr.py
11:24
<
arigo >
it's translated into a call to one of the helpers from rlib/*.py
12:25
lritter has joined #pypy
12:31
Rhy0lite has joined #pypy
13:09
petronny has joined #pypy
13:11
<
petronny >
Well, pypy3 building on rpi4 (8GB) is ok, but it's so close to OOM.
13:12
<
petronny >
rpython takes 64.8% of RAM, and there are 4 gcc processes and each one takes 5%.
13:16
<
mattip >
petronny: that is during the actual compilation via "make", after the python translator does all its work, correct?
13:17
<
petronny >
It seems that when gcc is running, rpython is doing nothing and just hold the 64.8% RAM.
13:18
<
mattip >
right, it could be cleaned out
13:18
<
petronny >
Great, but how?
13:19
<
petronny >
Oh, I'm not building with `make`.
13:19
<
mattip >
I played with it a long time ago when I needed to, but couldn't find what was holding the ref to the database object
13:21
<
mattip >
it was easier to just get more memory in my VM
13:24
<
mattip >
arigo: thanks. I think it is ll_decode_utf8 in rstr.py, which uses rstr.str_decode_utf8
13:25
<
cfbolz >
petronny: workaround: run rpython with --source, then run make at the end in the /tmp/usession*/testing_1 dir
13:28
<
petronny >
cfbolz: Thanks. I will give it a try.
13:30
<
petronny >
Do I need to run other commands after `make`?
13:34
<
cfbolz >
Hm, maybe, mattip, are the extra build steps put into make?
13:36
<
mattip >
nope. There is a copy of the artifacts back out of /tmp and then building the cffi extension libraries
13:43
YannickJadoul has joined #pypy
14:04
jcea has joined #pypy
14:08
<
petronny >
Here are the commands I'm going to run
14:08
<
petronny >
`python2 ../../rpython/bin/rpython -Ojit --shared targetpypystandalone --source`
14:08
<
petronny >
`make -C /tmp/usession*/testing_1`
14:08
<
petronny >
`PYTHONPATH=../.. ./pypy3-c ../../lib_pypy/tools/build_cffi_imports.py`
14:09
<
petronny >
Are these fine?
14:12
<
mattip >
I think you need to copy the pypy3-c and libpypy3.so from /tmp/usession*/testing_1 to pwd
14:13
<
mattip >
and when you run tools/build_cffi_imports.py, you need to cd to lib_pypy, since the modules are placed in pwd
14:14
speeder39_ has joined #pypy
14:27
<
petronny >
Thanks again. I'm running the first step now.
14:27
<
petronny >
And `--source` must be put before `targetpypystandalone`.
14:49
<
YannickJadoul >
mattip: wrt. the fix-fstring-double-braces branch: I know it's not pretty, but it's horrible to keep a balance between keeping it close to CPython (with its `ch = *s++` lines) and nicely-readable
15:20
YannickJadoul has quit [Quit: Leaving]
15:30
<
petronny >
BTW, does translating pypy not output those beautiful patterns any more?
16:24
speeder39_ has quit [Quit: Connection closed for inactivity]
16:56
Dejan has joined #pypy
16:56
<
mattip >
petronny: if you redirect stdout then no fractals
17:23
<
mattip >
I think PyUnicode_{En,De}codeLocale are the last things needed to build PyQT
17:24
<
mattip >
the next stage is to fix all the real problems with partially-implemented cpyext functions (hopefully there will not be many)
17:27
speeder39_ has joined #pypy
18:11
TheNewbie has quit [Quit: Leaving]
19:19
_whitelogger has joined #pypy
20:27
speeder39_ has quit [Quit: Connection closed for inactivity]
20:52
jcea has quit [Ping timeout: 256 seconds]
20:52
BPL has joined #pypy
21:20
<
mattip >
it seems this test somehow passes on the nightly linxu64 build on py3.6, but shouldn't
21:20
<
mattip >
python2 pytest.py pypy/objspace/std/test/test_stdobjspace.py -k test_wrap_string
21:20
<
mattip >
I get a BadUTF8 exception, the test wants a UnicodeDecodeError
22:06
Rhy0lite has quit [Ping timeout: 256 seconds]
22:50
lritter has quit [Quit: Leaving]
23:22
Rhy0lite has joined #pypy
23:36
oberstet has quit [Quit: Leaving]
23:44
jcea has joined #pypy