PyMol 0.92 has just been released on http://pymol.sf.net/. Please find attached the file "pymol-0.92.ebuild". The only change from the existing "app-sci/pymol-0.90.ebuild" is the removal of the line "epatch ${FILESDIR}/setup.py-gentoo.patch", since this patch is no longer necessary. With this ebuild, PyMol 0.92 compiles fine, at least on my Pentium-4 system. Reproducible: Always Steps to Reproduce: 1. 2. 3.
Created attachment 19843 [details] ebuild for PyMol 0.92
It has just been announced on pymol-users that some problems have been found in PyMol 0.92 and that a 0.93 bugfix version will be released ASAP.
Created attachment 20180 [details] ebuild for PyMol 0.93
I have attached pymol-0.93.ebuild for the app-sci category. It is exactly the same a pymol-0.92.ebuild except for its name and builds fine on a Pentium-4 system.
The 0.93 ebuild also works fine for me on an Athlon XP system
Could you please test pymol-0.95 that will be in portage later today? Thanks.
The 0.95 ebuild in the Portage tree installs correctly. When running PyMOL 0.95, I sometimes get a seqmentation fault during raytracing, but I also get that on non-Gentoo systems, so it's probably a PyMOL problem.
Maybe this is a gcc related problem, see http://sourceforge.net/tracker/index.php?func=detail&aid=806361&group_id=4546&atid=104546
I've been looking at the raytracing crash over the weekend. It doesn't happen when I compile PyMOL 0.95 "manually" (using python setup.py install), but it does occur when I use Portage with CFLAGS="-march=pentium4 -O2 -pipe -funroll-loops". After some elimination, it seems that the problem is caused by the "-fgcse" optimization flag enabled by "-O2". When I set CFLAGS (in /etc/make.conf) to "... -O2 -fno-gcse ...", the resulting PyMOL build doesn't crash. I was under the impression that compiler optimizations were not supposed to cause crashes :-), so it's probably a gcc bug. Anybody know how to report things like this?
Can you give some instructions on how to produce this crash? Maybe one could track it down to some specific gcc versions.... Maybe it is a gcc bug, maybe it just shows a bug in PyMol code, I think it's too early to tell. Re-opening, changing subject.
I created a small test script "crash.pml" which loads a small PDB file ("mgl.pdb"), creates a stick representation and starts the raytracing. I'll attach both files to the report. You can start PyMOL in non-GUI mode with "pymol -qc crash.pml". If the build is OK, it should print the time it took to make the image, otherwise it will segfault.
Created attachment 30587 [details] Structure file for PyMOL test script
Created attachment 30588 [details] PyMOL script to test raytracer crash Minimal script to trigger raytracer segfault
I have been testing 0.97 for a few days and it seems that this new version solves the ray-tracing bug. Aggressive optimizations no longer seem to make pymol segfault when raytracing. I you still have problems with the current version, please comment. Otherwise, I'll mark this bug as WONTFIX in a couple of weeks, and try to mark 0.97 stable as soon as possible.
Yup, 0.97 seems fine, even with "-fgcse" enabled. I'm still curious about the cause of the problem, but I don't really hav the time to delve into the necessary depths now :-). Thanks for the 0.97 ebuild!
I guess we'll never know if it was a gcc, PyMOL or X driver problem... But at least it works, now ;-)
It was the gcc. I have emerged 0.97 with gcc 3.3.2. I was having the same problem... then I upgraded the gcc to 3.3.3 and all is fine now..
