New submission from Christoph Reiter <reiter.christoph(a)gmail.com>:
For future reference (since I've given up for now):
- Checkout PyGObject git (I've tested 3.4/3.8, 3.4 has lower glib/gir
dependencies if that's a problem)
- Apply the following patches to make autotools work with PyPy:
- https://bugzilla.gnome.org/show_bug.cgi?id=696646
- https://bugzilla.gnome.org/show_bug.cgi?id=696648
- https://bugzilla.gnome.org/show_bug.cgi?id=696650
- Apply the attached patch that works around some limitations in cpyext.
- Run the following to build against PyPy (a nightly build for example):
PYTHON=/path_to_pypy/bin/pypy ./autogen.sh
make
- To run the tests:
make check
or better:
SKIP_PEP8=1 make check.gdb (since it segfaults and pep8 tests are slow)
The first segfault is because
baseclass.tp_new(subclass, ...) # baseclass != subclass
doesn't seem to work.
----------
files: pygobject-pypy-fix-v1.patch
messages: 5508
nosy: lazka, pypy-issue
priority: bug
status: unread
title: CPyExt errors with PyGObject
________________________________________
PyPy bug tracker <tracker(a)bugs.pypy.org>
<https://bugs.pypy.org/issue1434>
________________________________________
New submission from Christoph Reiter <reiter.christoph(a)gmail.com>:
The following crashes with the jit enabled (linux x64)
while 1: hash(tuple([0] + {"": 0}.keys()))
----------
messages: 5520
nosy: lazka, pypy-issue
priority: critical
status: unread
title: Crash with recent nightly builds
________________________________________
PyPy bug tracker <tracker(a)bugs.pypy.org>
<https://bugs.pypy.org/issue1435>
________________________________________
New submission from Brian Kearns <bdkearns(a)gmail.com>:
for list, performance looks not that different:
$ ./pypy/goal/pypy-c -m timeit -s "a = 'abc'" "'abc' in ('123', '456', 'abc',
'def')"
1000000000 loops, best of 3: 0.00107 usec per loop
$ ./pypy/goal/pypy-c -m timeit -s "a = 'abc'" "a in ('123', '456', 'abc',
'def')"
100000000 loops, best of 3: 0.00292 usec per loop
for set:
$ ./pypy/goal/pypy-c -m timeit -s "a = 'abc'" "'abc' in {'123', '456', 'abc',
'def'}"
1000000000 loops, best of 3: 0.00107 usec per loop
$ ./pypy/goal/pypy-c -m timeit -s "a = 'abc'" "a in {'123', '456', 'abc',
'def'}"
10000000 loops, best of 3: 0.025 usec per loop
looks like in the last case an optimization that is applied to list is not
applied to set?
----------
messages: 5521
nosy: bdk, pypy-issue
priority: performance bug
status: unread
title: missed optimization on set contains
________________________________________
PyPy bug tracker <tracker(a)bugs.pypy.org>
<https://bugs.pypy.org/issue1436>
________________________________________
New submission from Marcus Smith <qwcode(a)gmail.com>:
the user and home installation schemes are broke in pypy 1.9 due it's custom
"pypy" installation scheme.
user scheme: http://docs.python.org/dev/distutils/install.html#alternate-
installation-the-user-scheme
home scheme: http://docs.python.org/dev/distutils/install.html#alternate-
installation-the-home-scheme
there's a wake of issues/workarounds in pip and virtualenv related to the "pypy"
scheme:
-https://github.com/pypa/pip/issues/629
-https://github.com/pypa/pip/issues/624
-https://github.com/pypa/pip/issues/625
-https://github.com/pypa/virtualenv/issues/306
-https://github.com/pypa/virtualenv/issues/302
I created a diff that shows the difference for distutils.command.install between
py2.7 and pypy1.9, and discuss options.
https://github.com/qwcode/pypy_scheme/commit/b7856156f302c192698cd311df814d…
7a8a0
----------
messages: 4791
nosy: pypy-issue, qwcode
priority: bug
release: 1.9
status: unread
title: user and home install schemes broke in pypy
________________________________________
PyPy bug tracker <tracker(a)bugs.pypy.org>
<https://bugs.pypy.org/issue1272>
________________________________________
