On Feb 11, 2011, at 10:12 PM, Anthony wrote: > On Saturday, February 12, 2011 1:03:48 AM UTC-5, Jonathan Lundell wrote: > On Feb 10, 2011, at 6:19 AM, Massimo Di Pierro wrote: > > > > I offer $300 to whoever can identify within 3 weeks and without > > ambiguity the cause of this problem. The bounty applies even if the > > problem turns out to be not in web2py but in one of the Python > > modules, in Apache or in the mod_wsgi implementation. It does not > > apply if the problem is due to known bug that has already been fixed > > by the responsible party (for example if you are using an old un- > > patched python version like 2.5.0 or an old mod_wsgi version). > psycopg2 2.0.7 is the version in use in this case. 2.0.8 had this in the > change log: > > > * psycopg/pqpath.c (_pq_fetch_tuples): Don't call Python APIs > > without holding the GIL. > > > Suggestive, no? There's some evidence on the web that this may have led to > "Segmentation fault - premature end of script headers" errors. > > 2.3.2 is current > > Though someone reported the same problem with MySQL: > https://groups.google.com/d/msg/web2py/AcCDl5eJnds/AXNyKO6kmBwJ >
Part of the problem is that this is a very general error. Lots of things can cause a segmentation fault of the wsgi request thread, which in turn leads to that message. Psycopg2 2.0.7 may well have a problem that causes it, and there may well be other ways for it to happen as well. It's important to eliminate the known problems (like this one) so that if another fault comes along, we know it's something else. Unfortunately we can't carry these adapters along in web2py, since they're not pure/portable Python.
