Mailman 3 December 2013 - pypy-issue

[issue1328] cProfile: slow on built-in function calls
by Armin Rigo 01 May '14

01 May '14

New submission from Armin Rigo <armin.rigo(a)gmail.com>: The cProfile recording overhead with the JIT is low when counting function calls, but very high when counting calls to built-in functions. Jitviewer shows that it takes a lot of residual calls around a single call to a built-in function. It seems to me that we did not put enough efforts in optimizing this. A trivial example shows an overhead of about 500x: def f(n): lst = [] for i in xrange(n): lst.append(i) # built-in call lst.pop() # built-in call ---------- messages: 5010 nosy: arigo, pypy-issue priority: performance bug status: unread title: cProfile: slow on built-in function calls ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1328> ________________________________________

3 5

[issue1416] Fix ARM hard floats
by Fijal 28 Apr '14

28 Apr '14

New submission from Fijal <fijall(a)gmail.com>: Hard floats on ARM backend are broken. This essentially prevents PyPy from working on virtually any interesting ARM. We should fix it before 2.0 final ---------- messages: 5423 nosy: fijal, pypy-issue priority: critical status: unread title: Fix ARM hard floats ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1416> ________________________________________

2 2

[issue1448] pytest on default does not support multi-inherited classes
by James Lan 27 Apr '14

27 Apr '14

New submission from James Lan <jameslan(a)gmail.com>: On default branch, pytest is 2.2.4.dev2, which does not support multi-inherited classes, so that test cases for both jvm and cli backend can not run. On pytest-update branch, pytest is 2.3.5dev6, fixed this issue. Please merge it back to default. BTW, will it be better to use a released version of pytest instead of dev? ---------- messages: 5559 nosy: jameslan, pypy-issue priority: critical status: unread title: pytest on default does not support multi-inherited classes ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1448> ________________________________________

4 4

[issue1629] leak
by Glyph 27 Apr '14

27 Apr '14

New submission from Glyph <glyph(a)twistedmatrix.com>: I am lead to believe this has something to do with the JIT eventually tracing through an infinite multitude of possible loops in a long-running program with combinatorial code paths. Honestly I don't personally understand the issue, all I know is that twistedmatrix.com's deployment of trac on pypy leaks an infinite amount of memory and then crashes. My understanding is that arigo, at least, knows what's going on, because there's a branch working on addressing this. I hope that someone else will fill out better metadata. ---------- assignedto: arigo messages: 6295 nosy: arigo, glyph, pypy-issue priority: critical status: unread title: leak ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1629> ________________________________________

4 4

[issue1496] pypy 2.0.x crash (stack trace included)
by tonypypy 27 Apr '14

27 Apr '14

New submission from tonypypy <antoun.kanawati(a)gmail.com>: While trying out latest pypy, I got consistent core dumps on one program; here's the stack trace from 2.0.2; I had the same problem yesterday with 2.0.1, but didn't get a change to generate the trace. Pypy 2.0.2; built from source; installed with virtualenv: $ ./rpython/bin/rpython -Ojit pypy/goal/targetpypystandalone.py $ virtualenv -p $PWD/pypy-c /opt/pypy-2.0.2 $ cd ~/myproject $ . /opt/pypy-2.0.2/bin/activate (pypy-2.0.2)$ pypy bin/program.py < data.json ................................................................................ ... ................................................................................ ... ....................................................zsh: segmentation fault (core dumped) Trying again with /usr/bin/time: 2.22user 0.14system 0:03.03elapsed 77%CPU (0avgtext+0avgdata 159568maxresident)k 0inputs+0outputs (0major+53786minor)pagefaults 0swaps Each period is printed at the end of a function call that creates a few thousand sets, frozensets, tuples, lists. The program is purely sequential, and pure python (no C/C++ parts). I use cPickle instead of pickle. --------------------------------------------------------------------- No problem pypy 1.9. Haven't tried stock python (too slow for this). * Environment: Fedora core 17; 8GB RAM; intel i7 $ uname -a Linux lenovo-t420s 3.8.12-100.fc17.x86_64 #1 SMP Wed May 8 15:36:14 UTC 2013 x86_64 x86_64 x86_64 GNU/Linux # This pypy does not dump core. $ pypy -V Python 2.7.2 (341e1e3821fff77db3bb5cdb7a4851626298c44e, Jun 18 2012, 13:48:33) [PyPy 1.9.0 with GCC 4.7.0] ---------- messages: 5759 nosy: pypy-issue, tonypypy priority: critical status: unread title: pypy 2.0.x crash (stack trace included) ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1496> ________________________________________

4 4

[issue1445] Major regression since 2.0 beta
by Florian Apolloner 26 Apr '14

26 Apr '14

New submission from Florian Apolloner <florian(a)apolloner.eu>: Alex asked me to open the bug, I am myself not 100% sure it's pypy's fault -- but I am somewhat out of debugging options. Django's testsuite gets a lot of memory allocation errors under pypy like: ====================================================================== ERROR: test_complex_app (admin_scripts.tests.ManageValidate) manage.py validate does not raise an ImportError validating a complex app with nested calls to load_app ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/florian/sources/django.git/tests/admin_scripts/tests.py", line 1073, in test_complex_app out, err = self.run_manage(args) File "/home/florian/sources/django.git/tests/admin_scripts/tests.py", line 176, in run_manage return self.run_test('./manage.py', args, settings_file) File "/home/florian/sources/django.git/tests/admin_scripts/tests.py", line 140, in run_test universal_newlines=True).communicate() File "/home/florian/sources/pypy-c-jit-63155-972e776a277f-linux64/lib-python/2.7/subprocess.py", line 679, in __init__ errread, errwrite) File "/home/florian/sources/pypy-c-jit-63155-972e776a277f-linux64/lib-python/2.7/subprocess.py", line 1147, in _execute_child self.pid = os.fork() OSError: [Errno 12] Cannot allocate memory This is on a 64 bit machine with 4 Gig of RAM, once in a while even oom_killer hits it… I am also seeing segfaults on the ci-server, but those appear to be fixed with the recent nightlies. ---------- messages: 5551 nosy: apollo13, pypy-issue priority: critical status: unread title: Major regression since 2.0 beta ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1445> ________________________________________

4 5

[issue1564] sqlite3 can't import w/ OSX pypy-2.1-beta2
by mike bayer 26 Apr '14

26 Apr '14

New submission from mike bayer <mike_mp(a)zzzcomputing.com>: apologies if this is known, browsing the other sqlite3 issues they seem to apply to previous versions and aren't the same thing I'm getting, this is on OSX mountain lion, pypy build linked at https://bitbucket.org/pypy/pypy/downloads/pypy-2.1-beta2-osx64.tar.bz2 classics-MacBook-Pro:src classic$ pypy-2.1-beta2-osx64/bin/pypy Python 2.7.3 (fde1203a425d, Jul 25 2013, 10:37:04) [PyPy 2.1.0-beta2 with GCC 4.2.1 Compatible Apple LLVM 5.0 (clang-500.1.58)] on darwin Type "help", "copyright", "credits" or "license" for more information. And now for something completely different: ``<fijal> I love pypy'' >>>> import sqlite3 Undefined symbols for architecture x86_64: "_sqlite3_enable_load_extension", referenced from: __cffi_f_sqlite3_enable_load_extension in _cffi__g24366574x62c16045.o (maybe you meant: __cffi_f_sqlite3_enable_load_extension) ld: symbol(s) not found for architecture x86_64 collect2: ld returned 1 exit status Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/local/src/pypy-2.1-beta2-osx64/lib-python/2.7/sqlite3/__init__.py", line 24, in <module> from dbapi2 import * File "/usr/local/src/pypy-2.1-beta2-osx64/lib-python/2.7/sqlite3/dbapi2.py", line 27, in <module> from _sqlite3 import * File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/_sqlite3.py", line 273, in <module> """, libraries=['sqlite3'] File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/cffi/api.py", line 311, in verify lib = self.verifier.load_library() File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/cffi/verifier.py", line 68, in load_library self.compile_module() File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/cffi/verifier.py", line 56, in compile_module self._compile_module() File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/cffi/verifier.py", line 131, in _compile_module outputfilename = ffiplatform.compile(tmpdir, self.get_extension()) File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/cffi/ffiplatform.py", line 25, in compile outputfilename = _build(tmpdir, ext) File "/usr/local/src/pypy-2.1-beta2-osx64/lib_pypy/cffi/ffiplatform.py", line 50, in _build raise VerificationError('%s: %s' % (e.__class__.__name__, e)) VerificationError: LinkError: command 'gcc' failed with exit status 1 ---------- messages: 5997 nosy: pypy-issue, zzzeek priority: critical release: 2.1 status: unread title: sqlite3 can't import w/ OSX pypy-2.1-beta2 ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1564> ________________________________________

4 8

[issue1471] Re-enable strategies on py3k
by Philip Jenvey 25 Apr '14

25 Apr '14

New submission from Philip Jenvey <pjenvey(a)underboss.org>: Most of list/set/dict etc strategies were disabled on py3k as they became pretty broken w/ the big bytes/unicode switch. We also lacked an rpython unicode type strategy at the time, but we have that now These need to be re-enabled before the final release, obviously for the sake of performance Many of the strategy tests were disabled in 2db1e0a806e2, but there are also some others. I believe they are all py3k_skip's ---------- assignedto: pjenvey messages: 5654 nosy: pypy-issue priority: performance bug status: unread title: Re-enable strategies on py3k ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1471> ________________________________________

1 2

[issue1541] py3k: Re-enable rpython int sized app-level int
by Philip Jenvey 25 Apr '14

25 Apr '14

New submission from Philip Jenvey <pjenvey(a)underboss.org>: This is a significant performance degradation on the py3k branch, basically intobject.py was disabled due to various obscure breakages. I'm intending to kill int/long/bool's multimethods (unless someone else does first ;] ) on default before mucking around with these types again ---------- assignedto: pjenvey messages: 5938 nosy: pypy-issue priority: performance bug status: unread title: py3k: Re-enable rpython int sized app-level int ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue1541> ________________________________________

1 4

[issue818] MySQLdb-python 1.2.3 segmentation fault on PyPy 1.5 OSX 64bit
by Didip Kerabat 25 Apr '14

25 Apr '14

New submission from Didip Kerabat <didipk(a)gmail.com>: I'm not sure if this would help, but I documented the crash report here: https://gist.github.com/1120902 ---------- messages: 2901 nosy: didip, pypy-issue priority: bug status: unread title: MySQLdb-python 1.2.3 segmentation fault on PyPy 1.5 OSX 64bit ________________________________________ PyPy bug tracker <tracker(a)bugs.pypy.org> <https://bugs.pypy.org/issue818> ________________________________________

3 2