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
tayfun26 has quit [Read error: Connection reset by peer]
<kenaan>
arigo default 0dafeea23cbb /pypy/module/cpyext/: Issue #2797 Fix for PySet_Discard()
inhahe__ has joined #pypy
inhahe_ has quit [Ping timeout: 240 seconds]
<kenaan>
arigo default 0cb72dae7486 /lib-python/2.7/threading.py: Issue #2768 Fall back to not using _py3k_acquire(), in case eventlet patches various classes
illume has joined #pypy
fryguybob has quit [Remote host closed the connection]
lritter has joined #pypy
<mattip>
arigato: does issue 2768 need any fixes on py3.5?
<arigato>
mattip: no
<mattip>
thanks
<arigato>
just ignore this patch when merging from default to py3.5
jcea has joined #pypy
<mattip>
any more outstanding things for pypy6 release?
<arigato>
I'm looking at the recent issues now
<arigato>
#2752 is still open, right?
<mattip>
yes, I could not make progress with the issue2752 branch. Even a second call to PyObject_GetBuffer fails, (after the first one fails)
<mattip>
even though PyString_AsString succeeds
<arigato>
#2786: I'm about to commit at least some workaround
<mattip>
2752 - something is also off with the documentation of rffi.get_raw_address_of_string
<mattip>
there is one last linux32 own test failure - it is calling intmask(id(Class)) , but id(Class) is too big for int, it needs uint
<mattip>
yes. The error message is confusing since the implementation of gc.get_rpy_type_index calls rgc.get_rpy_type_index and raises that error for a negative result
<arigato>
ah yes
<arigato>
but that @not_rpython implementation is just a hack to get some number
<arigato>
the number doesn't really make sense before translation
<mattip>
so it could return long?
<arigato>
should be fixed like this:
<arigato>
...no, it should return an "int" but one that is not negative
<arigato>
after translation, it always returns a small positive int
ceridwen has quit [Ping timeout: 245 seconds]
<mattip>
this and db187a245708 make me scratch my head "how did it ever pass", but I guess they changed lately when the build slave was not really working
<arigato>
yes, likely
<mattip>
py3.5 is also shaping up nicely, very few failing own tests
<arigato>
I had a look at most failures of test_pypy_c at some point
<arigato>
it's a mixture of "just fix the test" and the occasional "well the jit could do better here"
ceridwen has joined #pypy
<mattip>
if I have some time I might just go through and fix them to document the current JIT state
speeder39 has quit [Quit: Connection closed for inactivity]
<arigato>
the point is more that if you fix the tests, we loose track of the cases where the jit should do better
<arigato>
maybe the corresponding tests should be @py.test.mark.fail() or however the current invocation of that looks like now
jaffachief has joined #pypy
antocuni has joined #pypy
<mattip>
antocuni: hi
<mattip>
arigato: something that came up on numpy about CPython - is there any way to know when gc.collect() called in a loop is "done"?
<mattip>
i.e. additional calls will do nothing. I suspect three calls should suffice, is there a case where it won't?
<antocuni>
mattip: hi
<mattip>
antocuni: I want to release soonish. Should I wait for gc-hooks?
<antocuni>
mattip: I think that gc.collect() always triggers a major collection, so it never happens that "it does nothing"
<antocuni>
although from point of view of the user, what happens is that there might be more __del__()s to call, so indeed gc.collect() might have observable effects. But I don't think you can know that in advance
<antocuni>
mattip: gc-hooks: feel free to release: there are still a couple of things I'd like to add before merging, like the ability to report the time spent in each gc section in milliseconds (right now it uses clock ticks)
<arigato>
gc.collect: there is no way to know, indeed. maybe we could hack to return the number of objects with a __del__ that have been found by this gc.collect() call
<antocuni>
moreover, about the release: am I wrong or rpython tests are NOT run/shown nightly?
<mattip>
ok, thanks. gc-hooks seems to provide some nice info, if you stub out some docstrings and give me hints I can help flesh them out
<mattip>
rpython tests are only run when there is a rpython change, and then the changes drift off the summary pages,
<antocuni>
ok. But then we need to be careful: e.g. rvmprof tests are failing on my machine because the C files are out of sync with the ones on github
<arigato>
hum, maybe the summary page shouldn't try to show these in the same section, then
<antocuni>
which is probably a release blocker
<arigato>
(unsure how to fix that actually)
<antocuni>
mattip: re gc-hooks: right now the 'stats' param which is passed to the hooks contains a 'duration' field which is expressed in ticks. I'd like to turn it into milliseconds before merging; I surely don't want to merge it before that, else we will break expectations as soon as I do the change
<mattip>
antocuni: about rvmprof failure, if you copy the files does it break other things?
<mattip>
I guess we could make a rpython summary page on the buildbot
<antocuni>
it depends on the nature of the change of course; I don't know what changed on github
<mattip>
ok, I thought maybe you tried it
<antocuni>
hopefully, if it's a small change it is enough to just copy the files and everything works
<antocuni>
no, I just remember that I saw a failure in rpython, but didn't check further
<antocuni>
to be fair, I think it's planrich's fault :-P. If he updates the files on github, he should do the same on pypy, I think
<mattip>
something like this seems to work to get all the rpython results
<mattip>
of course, buildbot is now at version 1.1, we are still using 0.8.8 and "pip install buildslave==0.8.8" fails, it seems pip now cannot understand the requirements file from long ago
<simpson>
So it looks like no matter what, whether the C type is `int` or `long long`, the Python value will be wedged into a box of that size, and if it doesn't fit, then OverflowError.
<simpson>
But it does handle the Python `long` for you, so you shouldn't have to worry.
demonimin has quit [Remote host closed the connection]
<solarjoe4>
I am using Python 3 anyway, so there is no long anymore, but as I read this, it does nothing, just return the value. But what if it is too large for C int?
<simpson>
Then OverflowError, I think.
<solarjoe4>
Oh, sorry, I missed that
demonimin has joined #pypy
<solarjoe4>
Ok, that clarifies it. Thanks!
energizer has quit [Ping timeout: 276 seconds]
mcyprian has joined #pypy
mcyprian has quit [Client Quit]
mcyprian has joined #pypy
<simpson>
No worries.
mcyprian has quit [Client Quit]
mcyprian1 has joined #pypy
jaffachief has quit [Quit: Connection closed for inactivity]
mcyprian1 has quit [Client Quit]
mcyprian has joined #pypy
tayfun26 has joined #pypy
antocuni has joined #pypy
mcyprian has quit [Quit: Leaving.]
mcyprian has joined #pypy
mcyprian has quit [Client Quit]
marky1991 has joined #pypy
marky1991 has quit [Ping timeout: 264 seconds]
jamesaxl has quit [Read error: Connection reset by peer]
jamesaxl has joined #pypy
raynold has quit [Quit: Connection closed for inactivity]
planrich has quit [Quit: Leaving]
solarjoe4 has quit [Quit: Leaving]
oberstet has quit [Ping timeout: 256 seconds]
lritter has quit [Remote host closed the connection]
irclogs_io_bot has quit [Remote host closed the connection]
lazka has joined #pypy
irclogs_io_bot has joined #pypy
marky1991 has quit [Ping timeout: 268 seconds]
marky1991 has joined #pypy
forgottenone has quit [Ping timeout: 265 seconds]
forgottenone has joined #pypy
jamesaxl has quit [Ping timeout: 240 seconds]
<antocuni>
mattip: did we send a PR to cython to solve the "type size changed" warning?
<antocuni>
I just compiled numpy with a nightly pypy and I still see it; I think it would be nice to solve it before doing the release, else users will be confused/upset and will report issues to us
illume has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
antocuni has quit [Ping timeout: 276 seconds]
jamesaxl has joined #pypy
drolando has joined #pypy
energizer has quit [Remote host closed the connection]
energizer has joined #pypy
oberstet has quit [Remote host closed the connection]