<simpson>
Ooh, nice. This is something I've been thinking about as I ponder adding weakmaps to Monte: How to ensure that testing doesn't need to touch the GC?
EWDurbin has quit [*.net *.split]
stillinbeta has quit [*.net *.split]
dustinm has quit [*.net *.split]
cadr has quit [*.net *.split]
alcarithemad has quit [*.net *.split]
__pv has quit [*.net *.split]
cfbolz has quit [*.net *.split]
EWDurbin_ is now known as EWDurbin
stillinbeta_ is now known as stillinbeta
cfbolz_ is now known as cfbolz
xcm has quit [Remote host closed the connection]
<njs>
(also, pdb has nothing to do with signals...?)
<njs>
simpson: I doubt it's possible to entirely prevent testing from touching GC, if your GC is user-visible at all (e.g. if you support weakrefs or destructors)
<njs>
and even then I guess some kinds of memory leak regression tests might still need to touch GC
xcm has joined #pypy
<simpson>
Yeah. Destructors/finalizers aren't likely to make it in as-is, but weakmaps are an important capability-security primitive that we need to do somehow.
<simpson>
RPython doesn't have ephemerons; that might be a way forward for a custom Monte runtime.
alcarithemad has joined #pypy
<simpson>
(Also you are right about pdb and using exceptions instead of signals. https://bpaste.net/show/87a049c5a09a works fine on PyPy. My mistake!)
<arigato>
I'm not sure I understand why cpython (2.7) doesn't print foo
<arigato>
ah my bad
<arigato>
a cycle containing a __del__ => not called
alawrence has joined #pypy
<arigato>
alawrence: pdb should work in any thread. If you can give more details about the "ValueError: signal only works in main thread" you're getting, maybe there is a bug in pypy somewhere
<mattip>
things get confusing sometimes when I put a set_trace() in a untranslated test, in the test itself (not in the rpython code)
<alawrence>
arigato: I created issue 3003 and stuck a description in there.
nunatak has joined #pypy
nunatak has quit [Client Quit]
marvin_ has quit [Remote host closed the connection]
marvin has joined #pypy
<kenaan>
andrewjlawrence winmultiprocessing 1140e7fc5df0 /lib_pypy/_winapi.py: Work in progress.
xcm has quit [Read error: Connection reset by peer]
<kenaan>
cfbolz py3.6 b4f7c179d353 /pypy/: fix issue #3001: weird codecs can return strange types when using codes.en/decode but not via unicode.encode/bytes.d...
xcm has quit [Remote host closed the connection]
xcm has joined #pypy
<kenaan>
arigo default 8fde6d376fef /pypy/module/array/: Don't raise ValueError when comparing arrays with out-of-bound unicodes
lritter has joined #pypy
<arigato>
cfbolz: fixed
<cfbolz>
cool
<kenaan>
arigo py3.6 4a70f02715ab /pypy/module/_codecs/: Fix. The general issue is space.newtext("string") where the string might contain some random byte-chars from app-level.
xcm has quit [Read error: Connection reset by peer]
<arigato>
shows how much used pdb really is: this bug is here at least from cpython 3.5, and it explodes as soon as you call pdb.set_trace() from a non-main thread after having called it from the main thread
<arigato>
a more positive interpretation would be that *threads* are not actually much used :-)
<cfbolz>
haha
<cfbolz>
yes, I am going with the latter too
<arigato>
a darker version is that as soon as you have threads, not even pdb can help you understand bugs
<cfbolz>
right
alawrence_ has joined #pypy
alawrence has quit [Ping timeout: 256 seconds]
nunatak has joined #pypy
nunatak has quit [Read error: Connection reset by peer]
<toad_poloesque>
I didn't know what it was without googling.
<toad_poloesque>
Even though now that I do know what it means "memory views of ctype structures" is a description of part of what PEP 3118 covers.
<toad_poloesque>
mattip: Also does this include the changes to the datetime C-API?
<toad_poloesque>
Or is that for a later release?
<mattip>
toad_poloesque: no, since that will change the binary API, it cannot be part of the 7.1 series
<toad_poloesque>
Ah, OK.
<mattip>
the delay does give a chance to implement the *Fold* api ...
<toad_poloesque>
Yeah, when's the ETA for that?
<toad_poloesque>
My highest priority right now is finishing my PyCon tutorial, but I can do the fold part after I'm done.
alawrence has joined #pypy
<alawrence>
arigato: It seems that someone reported this issue in 2011 https://bugs.python.org/issue13120. I am not sure what is going on there as it is down as being merged.
<alawrence>
and resolved...
<toad_poloesque>
alawrence: It was merged and resolved on March 9th and backported to Python 3.7
<toad_poloesque>
It would be in the latest alpha build of 3.8 and maybe 3.7.3, but not in 3.5
<mattip>
also not in 3.6?
<toad_poloesque>
I only see one backport.
<alawrence>
It was marked as being resolved in 2012
<toad_poloesque>
Oh hm, I was just looking at the merged PRs.
<mattip>
toad_poloesque: we probably won't do a 8.0 release for a few months
<toad_poloesque>
I guess the thing that was merged recently was a fix to one of the tests, my bad.
<toad_poloesque>
mattip: Cool, should have plenty of time to do the fold stuff by then.
marky1991 has joined #pypy
<mattip>
cool. Added the PEP 3118 link and published
oberstet has quit [Quit: Leaving]
xcm has quit [Ping timeout: 250 seconds]
xcm has joined #pypy
jacob22 has quit [Read error: Connection reset by peer]
<kenaan>
mattip buildbot fc95257ff7ab /bot2/pypybuildbot/builds.py: use fixed version of benchmark repo, forcing 'branch' or 'revision' refer to pypy Also use https:/speed.pypy.org...
bbot2 has quit [Quit: buildmaster reconfigured: bot disconnecting]
jcea has quit [Remote host closed the connection]
<kenaan>
mattip buildbot f4e9749d24ba /bot2/pypybuildbot/builds.py: fix argument order