Bugs item #1063937, was opened at 2004-11-10 17:50 Message generated for change (Comment added) made by birkenfeld You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1063937&group_id=5470
Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Python Interpreter Core Group: Python 2.2.3 Status: Open Resolution: None Priority: 7 Submitted By: munder12 (munder12) Assigned to: Nobody/Anonymous (nobody) Summary: Random core dumps Initial Comment: I have narrowed it down as far as I can by continuing to make the problem simpler and simpler but where it still core dumps. The way this is set up is the following: pytest2.py and pytest.py are in the same directory. pytest3.py is in PYTHONPATH where PYTHONPATH is <abs_path>/BASE:<abs_path>/SUP (pytest3.py is in BASE). Run ./pytest2.py several times. This current problem core dumps on average about 2 times every 5 runs. I have attached a file that has the Python listings as well as the gdb traceback. This is running under Fedora Core 1 with: Python 2.2.3 (#1, Oct 15 2003, 23:33:35) [GCC 3.3.1 20030930 (Red Hat Linux 3.3.1-6)] on linux2 Thank you, Mark PS Strangely enough the comments in pytest.py seem to actually increase the frequency of core dumps. ---------------------------------------------------------------------- >Comment By: Reinhold Birkenfeld (birkenfeld) Date: 2005-10-03 11:15 Message: Logged In: YES user_id=1188172 I'd say that it can be closed. As mwh says: "If Python 2.4b2 cored 25% of the time it was launched, someone else would have noticed by now :)" ---------------------------------------------------------------------- Comment By: Neal Norwitz (nnorwitz) Date: 2005-10-03 08:42 Message: Logged In: YES user_id=33168 I can't reproduce on current CVS (2.5). Can anyone reproduce this now? Should this be closed? ---------------------------------------------------------------------- Comment By: Nick Coghlan (ncoghlan) Date: 2004-11-23 10:15 Message: Logged In: YES user_id=1038590 I'm not sure it's relevant, but I once had a similar problem with Python flakiness that turned out to be due to some old .pyc files lying around (this was Fedora Core 3-test 1, and the offending files were in my Python CVS build directory). For whatever reason, Python wasn't picking up the version mismatch and was trying to use the old .pyc files. Seg faults abounded as a result. One thought: could root ownership of pre-generated .pyc's have that effect? (I don't know how Python reacts if it can't delete the .pyc's) ---------------------------------------------------------------------- Comment By: munder12 (munder12) Date: 2004-11-17 18:00 Message: Logged In: YES user_id=1156202 I have written a program in C that just opens and closes a file repeatedly. It appears to work fine. But, there appears to be much more that Python is doing behind the scenes than what my script is explicitly directing (open and close the file). Since I'm not sure what all OS related calls Python is making when opening, say, "site.py," I'm not quite sure how I can write a C code that mimics what Python is doing. It may well be that the OS is the culprit. However, it also could be that, in the Python code itself, some error checking is not being performed on all OS calls as they should be since they so rarely fail on a mjority of OS's. Or, extra try...catch blocks maybe could be added to retry the OS call(s) that is "incorrectly" failing on Fedora Core 1 so that Python maintains its portability with (hopefully) minimal speed impact. ---------------------------------------------------------------------- Comment By: Tim Peters (tim_one) Date: 2004-11-17 01:54 Message: Logged In: YES user_id=31435 At this point, it would be prudent to write the same kind of program in straight C, and test that. The more you find out, the less it appears that Python has something to do with what you're seeing. Note that it's not unusual to discover OS, compiler, and platform C library bugs by running Python programs, simply because Python builds on all of them. ---------------------------------------------------------------------- Comment By: munder12 (munder12) Date: 2004-11-17 01:47 Message: Logged In: YES user_id=1156202 It is 2 times out of 8 runs of the main script. Actually, that is 2 cores out of 1600 runs of the script that really cores. It does seem to be localized to Fedora Core 1. Fedora Core 2, Win 2000, Win XP, and Mandrake 9 on similar hardware do not have the problem with these scripts. The Python 2.4b2 is straight out of the tarball (compiled and installed cleanly). The core dumps occur randomly with posixpath.py, site.py, etc. and at different stages (robject() and within fstat() (from /usr/include/sys/stat.h)). ---------------------------------------------------------------------- Comment By: Michael Hudson (mwh) Date: 2004-11-16 09:38 Message: Logged In: YES user_id=6656 Hmm. The traceback shows that it's dumping core while loading the site.pyc file from the disk. This happens before Python begins executing whatever code you've asked it to. You could try deleting the site.pyc file and see if that makes any difference, or running python -S. I still think there must be something odd about your setup. If Python 2.4b2 cored 25% of the time it was launched, someone else would have noticed by now :) ---------------------------------------------------------------------- Comment By: munder12 (munder12) Date: 2004-11-11 17:05 Message: Logged In: YES user_id=1156202 I have checked out and built Python 2.4b2. It still core dumps on these same files but it now takes on average about 8 runs to get 2 cores. I will attach the 2.4b2 core traceback. ---------------------------------------------------------------------- Comment By: Michael Hudson (mwh) Date: 2004-11-11 16:34 Message: Logged In: YES user_id=6656 Uh, if that tiny thing cores, I'm very much inclined to believe there's something flaky in your setup. ---------------------------------------------------------------------- Comment By: munder12 (munder12) Date: 2004-11-11 16:25 Message: Logged In: YES user_id=1156202 I have played some more with the setup. Now only pytest2.py and pytest.py need to be present in same directory. Running pytest2.py will randomly cause core dumps coming from pytest.py. If you run pytest.py enough times directly it also will randomly core. I have attached pytest.py and pytest2.py. I have not tried with any other version of Python yet. ---------------------------------------------------------------------- Comment By: Michael Hudson (mwh) Date: 2004-11-11 15:47 Message: Logged In: YES user_id=6656 Have you tried a more recent Python? I'm finding decoding your report (attempting to reproduce) a little challenging. Could you upload a tarball containing a shell script driver or something? ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1063937&group_id=5470 _______________________________________________ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
