Antonio Cuni <anto.cuni(a)gmail.com> added the comment:
It seems that cpython behaves differently in the cases in which PYTHONPATH is
empty or unset:
anaconda foo $ echo $PYTHONPATH
:/home/antocuni/WebFormKit:/home/antocuni/pypy/trunk/:/home/antocuni/pypy/user/antocuni/:/home/antocuni/pypy/user/antocuni/hack/:/home/antocuni/pypy/misc/pg8000/:/home/antocuni/pypy/misc/casalino-carta/:/home/antocuni/pypy/misc/execnet/
anaconda foo $ python -c 'import site; print site'
<module 'site' from 'site.pyc'>
anaconda foo $ PYTHONPATH= python -c 'import site; print site'
<module 'site' from 'site.pyc'>
anaconda foo $ unset PYTHONPATH
anaconda foo $ python -c 'import site; print site'
<module 'site' from '/usr/lib/python2.6/site.pyc'>
I think the cause is that an empty PYTHONPATH triggers the insertion of two ''
in sys.path:
anaconda foo $ export PYTHONPATH=
anaconda foo $ python -S -c 'import sys; print sys.path'
['', '', '/usr/lib/python2.6/', '/usr/lib/python2.6/plat-linux2',
'/usr/lib/python2.6/lib-tk', '/usr/lib/python2.6/lib-old',
'/usr/lib/python2.6/lib-dynload']
anaconda foo $ unset PYTHONPATH
anaconda foo $ python -S -c 'import sys; print sys.path'
['', '/usr/lib/python2.6/', '/usr/lib/python2.6/plat-linux2',
'/usr/lib/python2.6/lib-tk', '/usr/lib/python2.6/lib-old',
'/usr/lib/python2.6/lib-dynload']
I think that this explains why regebro and armin see different results on
cpython. The problem in pypy still exists
_______________________________________________________
PyPy development tracker <pypy-dev-issue(a)codespeak.net>
<https://codespeak.net/issue/pypy-dev/issue505>
_______________________________________________________
Armin Rigo <arigo+pypyissues(a)codespeak.net> added the comment:
If it's related to the multijit branch merge, then the issue should disappear if
we don't actually use several JITs. Fijal, can you try to recompile a
pypy-c-jit with the jitdriver in pypy/module/__builtin__/functional.py disabled?
_______________________________________________________
PyPy development tracker <pypy-dev-issue(a)codespeak.net>
<https://codespeak.net/issue/pypy-dev/issue552>
_______________________________________________________
Armin Rigo <arigo+pypyissues(a)codespeak.net> added the comment:
How do I run the test so that it runs for 300 iterations and then stops?
----------
status: unread -> chatting
_______________________________________________________
PyPy development tracker <pypy-dev-issue(a)codespeak.net>
<https://codespeak.net/issue/pypy-dev/issue552>
_______________________________________________________
New submission from Fijal <fijall(a)gmail.com>:
When running twisted tests, there is an occasional crash. To reproduce: trial -u
twisted.names (if passes 300 iterations, stop and rerun, occurs very rarely).
Example of a gdb traceback. Repeated blackholes and metainterps on stack, crashes
with bad type id when collecting from optimizer.
----------
effort: ???
files: dump
messages: 1799
nosy: pypy-issue
priority: bug
release: ???
status: unread
title: crash in JIT
_______________________________________________________
PyPy development tracker <pypy-dev-issue(a)codespeak.net>
<https://codespeak.net/issue/pypy-dev/issue552>
_______________________________________________________
