Mailman 3 October 2016 - pypy-issue

Issue #2426: copy.copy performance regression (pypy/pypy)
by Maciej Fijalkowski 29 Oct '16

29 Oct '16

New issue 2426: copy.copy performance regression https://bitbucket.org/pypy/pypy/issues/2426/copycopy-performance-regression Maciej Fijalkowski: This code is ~2x slower on new pypy than on 4.0.1 (when run with 1000000). Possibly changes to how maps work? ``` #!python import copy, sys class A(object): def __init__(self, a, b, c): self.a = a self.b = b self.c = c self.d = a self.e = b a = [A(1, 2, 3), A(4, 5, 6), A(6, 7, 8)] def f(): for i in range(int(sys.argv[1])): copy.copy(a[i % 3]) f() ```

1 0

Issue #2425: str error from 'future' module in PyPy2 (pypy/pypy)
by Daniil Yarancev 26 Oct '16

26 Oct '16

New issue 2425: str error from 'future' module in PyPy2 https://bitbucket.org/pypy/pypy/issues/2425/str-error-from-future-module-in… Daniil Yarancev: ## Error ##: ``` #!python >>>> print(str('ghbdtn')) Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/home/tiber/bench/pypy2-v5.4.1-linux64/site-packages/future/types/newstr.py", line 204, in encode if errors == 'surrogateescape': UnboundLocalError: local variable 'errors' referenced before assignment ``` ## How to reproduce: ## ``` #!bash pypy -m pip install future ``` And try this code: ``` #!python from __future__ import unicode_literals from builtins import str print(str('hello')) ```

1 0

Issue #2424: pypy-dev is required to install numpy from head (pypy/pypy)
by lesshaste 21 Oct '16

21 Oct '16

New issue 2424: pypy-dev is required to install numpy from head https://bitbucket.org/pypy/pypy/issues/2424/pypy-dev-is-required-to-install… lesshaste: In the docs at http://pypy.org/download.html#installing-numpy and http://doc.pypy.org/en/latest/install.html steps are set out for installing pypy+numpy . In order for pip install git+https://github.com/numpy/numpy.git to work in ubuntu you need to first install pypy-dev however. This is not mentioned in the docs presently.

1 0

Issue #2423: [cpyext] wrong __new__ assigned to tp_new for app-level classes (pypy/pypy)
by mattip 20 Oct '16

20 Oct '16

New issue 2423: [cpyext] wrong __new__ assigned to tp_new for app-level classes https://bitbucket.org/pypy/pypy/issues/2423/cpyext-wrong-__new__-assigned-t… mattip: I created a test in the ``missing-tp_new`` branch that demonstrates how an app-level class's ``__new__`` is not assigned to the ``tp_new`` slot, rather the generic ``object``'s ``__new__`` is assigned instead. AFAICT the problem is in the use of ``typedef = w_type.layout.typedef`` in line 249 of cpyext.py. The typedef is not unique to the w_type, and this non-unique object is used as a part of the key to lookup wrappers in the slotdef.py ``SLOTS`` dictionary. I think this is only a problem for the ``__new__`` method. I tried working around it by creating another ``NEW`` dictionary for the tp_new slot (as a special case in ``update_all_slots``, but could not get the pieces quite in the right place to make it translate.

1 0

Issue #2422: Unexpected inversion of performance between objects and tuple (pypy/pypy)
by Pierre-Antoine Champin 19 Oct '16

19 Oct '16

New issue 2422: Unexpected inversion of performance between objects and tuple https://bitbucket.org/pypy/pypy/issues/2422/unexpected-inversion-of-perform… Pierre-Antoine Champin: Consider a simple object type made of a few values (e.g. geometrical point). It is possible to implement it with a class, or to use tuples. There are also intermediate solutions, such as a class with __slots__ or a namedtuple. I was wondering what were the implications, in terms of performance, of those different options. My intuition was that, when it comes to comparison, tuples would be more efficient because their comparison operators are built-in, while for objects I needed to implement them with special methods. So I wrote the attached test-case (also available [here](https://gist.github.com/pchampin/38e2a4ac66d6baea44f17a0ce3318293)). It confirmed my intuition: with CPython (2 and 3), comparing tuples (including named tuples) is 10x faster than comparing objects with customized __eq__ and __lt__ methods. However, with Pypy (5.1.1), the results are totally opposite: object comparison are 10 to 20 times faster than tuple comparison, which take roughly the same time in Pypy as in CPython! This is all he more surprising that, in order to implement __eq__ and __lt__ for my classes, I just store their attributes in tuples and compare the two tuples... So why would this be slower when I directly compare tuples ?? It seems to me that Pypy could use the same optimization in both places, and that there might be, if not a bug, at least a missed opportunity here -- although my understanding of the inners of Pypy is admittedly limited..

1 0

Issue #2421: multiprocessing.Pool under PyPy3.5 fails (pypy/pypy)
by Daniil Yarancev 19 Oct '16

19 Oct '16

New issue 2421: multiprocessing.Pool under PyPy3.5 fails https://bitbucket.org/pypy/pypy/issues/2421/multiprocessingpool-under-pypy3… Daniil Yarancev: Trying this simple example with PyPy3.5: ``` #!python from multiprocessing import Pool def f(x): return x*x if __name__ == '__main__': p = Pool(5) print(p.map(f, [1, 2, 3])) ``` CPython 3.5.2 works as expected: [1, 4, 9] PyPy3.5 fails with error: https://paste.pound-python.org/show/IoKfW0weIXMImtbKYvUO/

1 0

Issue #2420: [patch] FreeBSD requires sys/ttycom.h for struct winsize (pypy/pypy)
by David Naylor 14 Oct '16

14 Oct '16

New issue 2420: [patch] FreeBSD requires sys/ttycom.h for struct winsize https://bitbucket.org/pypy/pypy/issues/2420/patch-freebsd-requires-sys-ttyc… David Naylor: FreeBSD defines struct winsize in sys/ttycom.h. The attached patch adds the required include for PyPy3 to discover struct winsize.

1 0

Issue #2419: [patch] FreeBSD has depreciated the use of sys/timeb.h (pypy/pypy)
by David Naylor 14 Oct '16

14 Oct '16

New issue 2419: [patch] FreeBSD has depreciated the use of sys/timeb.h https://bitbucket.org/pypy/pypy/issues/2419/patch-freebsd-has-depreciated-t… David Naylor: The attached patch adds FreeBSD (with OpenBSD) to the list of platforms that do not use "sys/timeb.h". Due to this ftime is obsoleted and a fix to interp_time is needed to avoid importing undefined values TIMEB and c_ftime from rpython.rlib.rtime.

1 0

Issue #2418: [patch] pypy3-5.5.0 does not compile on PowerPC (pypy/pypy)
by David Naylor 14 Oct '16

14 Oct '16

New issue 2418: [patch] pypy3-5.5.0 does not compile on PowerPC https://bitbucket.org/pypy/pypy/issues/2418/patch-pypy3-550-does-not-compil… David Naylor: Following on from Issue #2369, the latest release of pypy3 (5.5.0) does not compile on PowerPC without the attached patches.

1 0

Issue #2417: segmentation fault when trying to connect on a closed socket with AsyncIO (pypy/pypy)
by Ludovic Gasc 14 Oct '16

14 Oct '16

New issue 2417: segmentation fault when trying to connect on a closed socket with AsyncIO https://bitbucket.org/pypy/pypy/issues/2417/segmentation-fault-when-trying-… Ludovic Gasc: Hi, I'm using pypy-c-jit-87739-57696c7e37eb-linux64, nightly version of pypy3.5. If I launch this file without a server that listens on 8000: https://github.com/asyncio-doc/asyncio-doc/blob/master/examples/aiohttp_cli… I've this error: /home/lg/Softwares/pypy-c-jit-87739-57696c7e37eb-linux64/lib-python/3/importlib/_bootstrap.py:223: UserWarning: compiletime version 3.3 of module 'aiohttp._websocket' does not match runtime version 3.5 return f(*args, **kwds) Warning: uncaught exception in callback: bootstrap <RThreadError object at 0x7f87f30ebac8> Traceback (most recent call last): Erreur de segmentation (core dumped) FYI, I'm at PyCON-FR with Ronan, I'm trying to test PyPy3.5 with several AsyncIO snippets. For now, most things works as expected, except ssl module. Have a nice day.

1 0