speeder39_ has quit [Quit: Connection closed for inactivity]
<arigato>
cfbolz, ronan: +1 for an rpython utf-8 type. Some design is indeed needed; e.g. we need to think if slicing an utf-8 string returns an utf-8 string (then it needs to compute its shorter length) or a plain string or just be disallowed and replaced with our own methods
<kenaan>
lunixbochs Ryan-Hileman/add-support-for-zipfile-stdlib-1562420744699 870bee284470 /pypy/module/sys/initpath.py: add support for zipfile stdlib
<mattip>
which reminds me - we never switched to the new website ...
<Dejan>
I am making the following mistake every now and then: I make a venv with /opt/pypy3.6/bin/pypy3 -m venv tstenv
<Dejan>
i activate the environment
<Dejan>
and then i try to use pypy3
<Dejan>
problem is that venv does not create the pypy3 symlink :)
<Dejan>
so it uses Fedora's pypy3 :)
<Dejan>
in /usr/bin
<Dejan>
which is quite old
<Dejan>
the venv creates pypy3.6 symblink only
<Dejan>
and ofc, python and python3 symlinks
<Dejan>
so question is - is this something that should be changed upstream?
tsaka__ has joined #pypy
<mattip>
venv is in the stdlib (lib-python/3/venv) so it is shipped with pypy and is pure python, you should be able to find where it is making those symlinks
<mattip>
pull request welcome :)
<Dejan>
mattip, my question is - should i submit a PR in pypy or in python?
<Dejan>
is there something like git-svn for mercurial? :)
<mattip>
well, theoretically both, but the upstream fix would only percolate back when we resync. We should fix it in pypy sooner
<Dejan>
ok, thanks mate
<Dejan>
I will give it a try
<mattip>
if you don't find a way to use bitbucket and mercurial, put it on a pastebin and link it here, or send a diff to the pypy dev list
BPL has joined #pypy
BPL has quit [Quit: Leaving]
<Dejan>
mattip, the default branch has no lib-python/3
<mattip>
you want to fix pypy3, that is on the py3.6 branch
<mattip>
"hg update py3.6
<mattip>
"
<Dejan>
yea, found that already, thanks
<Dejan>
i use mercurial extremely rare once a year or so
<Dejan>
mattip, do you have the new website up and running somewhere?
<mattip>
not really
<Dejan>
how difficult it is to switch to the new website? :)
<mattip>
we want a site we can easily update, this one is written directly in html. We want that to be the output of some generate-from-templating system
<mattip>
so start from text, then use some hugo/jinja2/??? thingy to get the html generated
lordmauve_ has left #pypy [#pypy]
<mattip>
otherwise updating the site becomes a major task
xcm has quit [Read error: Connection reset by peer]
xcm has joined #pypy
<plluksie>
Hello. It seems cffi 1.12.3 doesn't like gcc 8.3.0 on smartos (illumos/solaris). I do wonder if it is specific to gcc or illumos. On gcc7 all is fine. https://pastebin.com/789EaRcz
fryguybob has quit [Quit: leaving]
<kenaan>
rlamy py3.7 0354a250d371 /pypy/module/: Update Python version numbers
forgottenone has joined #pypy
<mattip>
plluksie: can you compile any c-extension module? The message says (line 32) "Likely, the problem is not related to CFFI but generic to the setup.py of any Python package that tries to compile C code"
<mattip>
setuptools/distutils gets a bunch of CFLAGS from the host python via
<mattip>
sysconfig.get_config_var('CFLAGS')
<mattip>
among others
<mattip>
I doubt msave-args is coming from cffi
dddddd has joined #pypy
[Arfrever] has quit [Ping timeout: 245 seconds]
[Arfrever] has joined #pypy
<mattip>
the py3.6 crash with aarch64 is on rlib.rsiphash.ll_hash_string_siphash24
<mattip>
in a graph where it is called with a rpy_unicodePtr
kanaka has quit [Ping timeout: 258 seconds]
kanaka has joined #pypy
<kenaan>
rlamy py3.7 c252447e48e9 /: Update stdlib to 3.7.4
<kenaan>
rlamy py3.7 2d29adb1a8b7 /pypy/tool/dis3.py: Make pypy/tool/dis3.py valid Python2
<kenaan>
rlamy py3.7 a0c09c770864 /lib-python/3/opcode.py: Use pypy3.6 opcodes for now
<kenaan>
rlamy py3.7 db91b0c1a492 /pypy/config/pypyoption.py: _thread is required by importlib now
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
abrown has quit [Quit: Leaving]
antocuni has joined #pypy
<arigato>
mattip: back now
<arigato>
thanks for answering the cffi question
<mattip>
feel like looking at the screen on arm64-2 ?
<mattip>
I am in there now but can move aside
<mattip>
I am bumbling around trying to figure out why the false positive
<mattip>
s/am/was/
<arigato>
looking... (sorry if I kicked you out the screen or something)
<mattip>
no problem, hopefully you can figure it out
<arigato>
try "screen -x"
<mattip>
maybe some kind of hash collision on the seen cache?
<mattip>
I ran a translation on x64 with a "if func.func_name == 'll_hash_string_siphash24': pdb.set_trace()" in transform_graph
xcm has quit [Remote host closed the connection]
<arigato>
do you know if it only occurs on aarch64?
<arigato>
or we've just seen it once on aarch64 I guess?
<mattip>
and stopped on the same graph (the second time through the function with the rpy_unicodePtr)
<arigato>
ah, it may just be the "range()" in here
<mattip>
it repeats every time on aarch64 when setting isatty to return False
xcm has joined #pypy
<mattip>
hehe, "just"
<arigato>
scroll back a couple more pages to see prints showing where the can-collect operation is
<arigato>
...indeed it's ll_newrange
<arigato>
no clue at all about why it shows up now in this case, but not in others
<mattip>
ahh, right, so the last thing the CollectAnalyzer prints
<arigato>
in this case yes, but in general it's a call stack
<arigato>
and we have to see where in the call stack there's the problematic call
<mattip>
I have the graph of the same function on x64, and on both platforms did xxx = [x.operation for x in graph.iterblocks()]
<mattip>
on x64 it is 10 operations, I think the aarch64 has 13 operations
<mattip>
sorry if that is just noise
<arigato>
I have no idea
<kenaan>
rlamy py3.7 a10d1ef767e8 /pypy/module/_frozen_importlib/moduledef.py: importlib._bootstrap._install() was split into 2 funcs due to PEP432/587 (bpo-22257)
<arigato>
in theory the malloc-remover should get rid of that range()
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
<cfbolz>
arigato: sorry, that was too funny to pass up, but feel free to veto this topic
<arigato>
I won't :-)
<cfbolz>
:-)
<cfbolz>
hm, should we make our own LOOKUP_METHOD/CALL_METHOD work exactly like LOAD_METHOD/CALL_METHOD in cpy3.7?
<cfbolz>
or is the behaviour the same?
<arigato>
worth checking, probably
<mattip>
it is strange that rpython.tools.ansiprint.isatty can change translation on aarch64
<mattip>
as if there is some race condition somewhere
antocuni has quit [Ping timeout: 245 seconds]
<mattip>
I moved the screen session out of the buildslave directory to a pypy checkout, added a bunch of checkgraph calls to backendopt.all, set isatty to return False, and started a translation
* mattip
zzz
<mattip>
note to self: should look at the macOS _ssl build failures
Rhy0lite has quit [Quit: Leaving]
marky1991 has quit [Remote host closed the connection]