03:45 <njs> regarding the logging lock thing: there is this old issue that might be relevant, since I guess pypy uses the python implementation of RLock: https://bugs.python.org/issue13697

03:48 <njs> ...actually, wait, does pypy really use the python implementation of RLock? trio definitelky depends on threading.RLock being implemented in C/RPython :-(

04:11 <njs> (update: pypy3 is safe, trio is safe, and I strongly suspect that this is in fact the issue that Pillow and dateutil are hitting, see my comments on https://bitbucket.org/pypy/pypy/issues/2905/ for details)

07:35 <kenaan> arigo default a12ac2eaf1d6 /pypy/module/pwd/interp_pwd.py: issue #2840 Don't release the GIL in the pwd module

07:55 <kenaan> arigo cffi/cffi b50894d36bae /testing/embedding/: #383 Increase the sleep time in this test

09:23 <kenaan> arigo py3.5 1d9db7971be3 /pypy/module/thread/os_lock.py: Forward-port from default

09:30 <kenaan> arigo rlock-in-rpython 2cb3f4e48db6 /: Issue #2905 A branch to backport py3.5's RPython version of W_RLock

09:30 <kenaan> arigo py3.5 a70dd232b6a3 /pypy/module/thread/test/test_lock.py: Don't stick these tests into some unrelated test class, particularly one that is skipped on some platforms

09:55 <kenaan> arigo py3.5 78890f8b8979 /pypy/module/thread/os_lock.py: aaaaaaaaa

10:04 <kenaan> arigo default b01d1db30165 /pypy/module/cpyext/test/test_misc.py: fix test

10:15 <kenaan> arigo rlock-in-rpython 4874b6c33f95 /: Backport

10:21 <kenaan> arigo rlock-in-rpython 8f94cecf5830 /lib-python/2.7/threading.py: fix

16:43 <kenaan> arigo rlock-in-rpython 4f83e855f001 /: close branch, ready to merge

16:43 <kenaan> arigo default 55e3c71a37fa /: hg merge rlock-in-rpython

23:27 <mssun> Hi, I'm using RPython to write code. Is it possible to call PyPy to interpret some Python code with RPython code? I'm trying to use CFFI, but some functions (symbols) of libpypy and RPython compiled code are redefined.

23:31 <tos9> mssun: Why are you writing code in RPython?

23:31 <mssun> To speedup my code actually.

23:32 <tos9> mssun: Python code?

23:32 <tos9> mssun: RPython isn't really a language you use to write code

23:32 <tos9> If you want it faster, typically you just run it in PyPy

23:32 <tos9> mssun: RPython is for writing interpreters

23:33 <tos9> (Not to mention AFAIK CFFI is not valid RPython. There's rffi, but again, not really a thing you'd use to write a regular ol' program)

23:34 <mssun> Yes, I know that RPython is mainly designed for writing interpreter.

23:37 <mssun> Sorry, I mean, use rffi in RPython to call cffi to interpret Python.

23:43 <mssun> Apparently, it doesn't work. I know that PyPy and RPython should not be used like this. The performance gain of RPython is great, so I'm thinking to use it to write some performance-sensitive code. However, we also have needs to call PyPy to interpret/JIT some Python code. (I probably need to hack PyPy)