trace987 has quit [Remote host closed the connection]
yizawa has quit [Quit: Connection closed for inactivity]
stkrdknmibalz has joined #pypy
commandoline has quit [*.net *.split]
tumbleweed has quit [*.net *.split]
commandoline has joined #pypy
tumbleweed has joined #pypy
tumbleweed has joined #pypy
tumbleweed has quit [Changing host]
energizer has quit [*.net *.split]
Lightsword has quit [*.net *.split]
cfbolz has quit [*.net *.split]
amirouche has quit [*.net *.split]
cfbolz has joined #pypy
amirouche has joined #pypy
Lightsword has joined #pypy
energizer has joined #pypy
amirouche has quit [Quit: WeeChat 2.8]
amirouche has joined #pypy
LarstiQ has joined #pypy
<fijal>
hi
<fijal>
arigato: ping
<fijal>
cfbolz: hey
<fijal>
we have not decided what's the best way to go about a decorator right?
<cfbolz>
Nope
<fijal>
do we have a set of cffi helpers somewhere?
<fijal>
like a special cffi module that's just helpers
<fijal>
if not, I'll add it
* fijal
tries with openssl first
<fijal>
what's the current story? do you develop on default, then port to py3?
slav0nic has joined #pypy
slav0nic has quit [Remote host closed the connection]
slav0nic has joined #pypy
<fijal>
ok what's the procedure of getting this branch into py3.7 (or py3.8)?
<fijal>
(I just pushed it I guess)
<fijal>
I kicked some buildbots, but they're offline?
<fijal>
cfbolz: I gonna need some help getting this in, maybe a review?
<cfbolz>
fijal: happy to review
<cfbolz>
The buildbots are offline?!
<fijal>
can you have a look? branch is called scoped-cffi-malloc
<fijal>
some of them
<cfbolz>
fijal: there were builds yesterday
<fijal>
yeah
<cfbolz>
fijal: the helper needs unit tests
<cfbolz>
Also, _pypy_util is very generic, maybe adding _cffi is indeed good
<fijal>
yeah well, it adds a malloc and a free, that's very generic
<fijal>
how do you test "did not segfault"? cause that's a bit of a test, but I guess I can write a test for whether it does the same thing as ffi.new
<fijal>
cfbolz: wrote a dumb test, it's not clear to me how to do better (you can do better with all the ifs for sizeof, but I gonna wait for armin to comment)
<cfbolz>
fijal: I meant the if logic, yes
<fijal>
how do I make a random struct?
<fijal>
cfbolz: yeah, maybe
<fijal>
can I merge it?
<cfbolz>
fijal: did you run the buildbot?
<fijal>
I tried, buildslave is off for linux-64
<cfbolz>
fijal: imo add some tests for the if-cases or remove the if-cases. and rename the file to have cffi in the name
<fijal>
I added one for the if case, I'm not sure if other if case is possible?
<fijal>
I would like to ask armin how to calculate an input to malloc
<fijal>
this is quite dumb, like I don't know how to test anything at all (seems to work?)
<fijal>
this is quite annoying, can I make branches of py3.7?
<cfbolz>
yes
<fijal>
because if not, it means I have to merge something to default before testing it
<fijal>
ok
<fijal>
so well, this time I'll just merge it and I'll make it off 3.7 next time
<fijal>
cfbolz: ok, so there is already _pypy_util_cffi
<fijal>
do I rename it to _pypy_util_cffi_inner?
<fijal>
it's a bit of a mess to just expose malloc/free I think
yizawa has joined #pypy
<cfbolz>
ah, I see
<cfbolz>
it's the one that's generated by cffi?
jacob22_ has quit [Ping timeout: 265 seconds]
fotis has joined #pypy
<fijal>
yeah
<fijal>
I did rename it, makes sense though
<fijal>
can you look now if you happy?
<cfbolz>
fijal: did you push?
<fijal>
pushing
<fijal>
I'm going to get a fresh pypy-c, so it tests the build process and then merge if all works nicely
<cfbolz>
cool
<cfbolz>
fijal: just imagine that we used to do all development that way 😱
fotis has quit [Ping timeout: 265 seconds]
fotis has joined #pypy
stkrdknmibalz has quit [Quit: WeeChat 3.0.1]
<fijal>
I don't understand how it works
<fijal>
why can I not run the build_cffi_modules? it can't find paths at all, no matter where I run it (which makes me quite skeptical, why does it have to run in lib_pypy? does that work at all?)
<arigato>
Corbin: not completely sure, but yes, just use "hg update open-ended-traces"
fotis has quit [Ping timeout: 245 seconds]
<cfbolz>
arigato, Corbin: yes exactly
<Corbin>
For reasons, I ended up using the release-candidate branch. (I'm on Nix, so it's easier to just pin a version corresponding to a release candidate than to hack a development branch.)
<Corbin>
I'm waiting for a fresh JIT to compile now, but I am hopeful because Monte has historically had massive traces. We do so much runtime audition of code.
yizawa has joined #pypy
fotis has joined #pypy
<cfbolz>
Corbin: the 3.8 release branch?
<Corbin>
cfbolz: Yeah. It looks like building the Monte stdlib got about 10% faster, and a long-running raytracer got about 21% faster.
<Corbin>
That's a pretty big impact given that I just changed a few build hashes and fixed up a couple build configuration issues.
<cfbolz>
scary 😅
fotis has quit [Ping timeout: 252 seconds]
fotis has joined #pypy
fotis has quit [Ping timeout: 245 seconds]
yizawa has quit [Quit: Connection closed for inactivity]