M
Mike C. Fletcher
There was a post a few months ago from "snacktime" in which they
described a failure when running under Twisted:
http://mail.python.org/pipermail/python-list/2005-February/265137.html
which they eventually narrowed down to an apparent problem with pycrypto:
http://mail.python.org/pipermail/python-list/2005-February/265146.html
and which generated this sourceforge bug report for Python:
http://twistedmatrix.com/pipermail/twisted-python/2005-February/009475.html
http://sourceforge.net/tracker/index.php?func=detail&aid=1080660&group_id=5470&atid=105470
I am seeing a similar effect, (though only after the program has been
running a few days), in one of our products running on FreeBSD 5.3-RC2
with Python 2.3.4, PyOpenSSL 0.6, Twisted 2.0, and PyPgSQL 2.4 (which is
somewhat different than snacktime's library collection (different
OpenSSL and PostgreSQL bindings)). Note particularly, that the PyCrypto
library to which the problem was apparently localised is *not* installed
on the failing machine.
As with snacktime, I see the failure coming from pthread_testcancel()
(gdb) bt
#0 0x28222f17 in pthread_testcancel () from /usr/lib/libpthread.so.1
#1 0x2821b39f in pthread_mutexattr_init () from
/usr/lib/libpthread.so.1
#2 0x00000000 in ?? ()
I haven't been able to isolate the failures to a particular piece of
code (this is showing up on a live server, so I'm hesitant to run under
gdb for a couple of days to try to catch the failure). I'm *not* using
RSA encryption (which was the area to which snacktime localised their
problem). I *do* use SSL sockets from Twisted, however. I am running a
background thread for database access, with a Queue mediating the set of
items to process and reactor.callFromThread returning control to Twisted
when the items are finished.
Anyway, my question is really "has this been solved already"? If so,
what was the solution? If not, I'll have to work at building a minimal
failing test case so I can begin to debug it.
Thanks for any information,
Mike
--
________________________________________________
Mike C. Fletcher
Designer, VR Plumber, Coder
http://www.vrplumber.com
http://blog.vrplumber.com
described a failure when running under Twisted:
http://mail.python.org/pipermail/python-list/2005-February/265137.html
which they eventually narrowed down to an apparent problem with pycrypto:
http://mail.python.org/pipermail/python-list/2005-February/265146.html
and which generated this sourceforge bug report for Python:
http://twistedmatrix.com/pipermail/twisted-python/2005-February/009475.html
http://sourceforge.net/tracker/index.php?func=detail&aid=1080660&group_id=5470&atid=105470
I am seeing a similar effect, (though only after the program has been
running a few days), in one of our products running on FreeBSD 5.3-RC2
with Python 2.3.4, PyOpenSSL 0.6, Twisted 2.0, and PyPgSQL 2.4 (which is
somewhat different than snacktime's library collection (different
OpenSSL and PostgreSQL bindings)). Note particularly, that the PyCrypto
library to which the problem was apparently localised is *not* installed
on the failing machine.
As with snacktime, I see the failure coming from pthread_testcancel()
(gdb) bt
#0 0x28222f17 in pthread_testcancel () from /usr/lib/libpthread.so.1
#1 0x2821b39f in pthread_mutexattr_init () from
/usr/lib/libpthread.so.1
#2 0x00000000 in ?? ()
I haven't been able to isolate the failures to a particular piece of
code (this is showing up on a live server, so I'm hesitant to run under
gdb for a couple of days to try to catch the failure). I'm *not* using
RSA encryption (which was the area to which snacktime localised their
problem). I *do* use SSL sockets from Twisted, however. I am running a
background thread for database access, with a Queue mediating the set of
items to process and reactor.callFromThread returning control to Twisted
when the items are finished.
Anyway, my question is really "has this been solved already"? If so,
what was the solution? If not, I'll have to work at building a minimal
failing test case so I can begin to debug it.
Thanks for any information,
Mike
--
________________________________________________
Mike C. Fletcher
Designer, VR Plumber, Coder
http://www.vrplumber.com
http://blog.vrplumber.com