GNOME Bugzilla – Bug 139447
crash -- thumbnail problem
Last modified: 2004-12-22 21:47:04 UTC
Here are the two PDFs that have caused the trouble: http://www.freedomhouse.org/research/mrr2004.pdf http://www.freedomhouse.org/pdf_docs/research/freeworld/2003/map2003.pdf 1.) Have Mozilla-Firefox set to open *.pdf with gnome-pdf-viewer 2.) Open either of the previous documents, and have Firefox use gnome-pdf-viewer to open them 3.) Open completion and opening, gnome-pdf-viewer will crash Here are two associated stack/backtraces from the previous files: Backtrace was generated from '/usr/libexec/gnome-pdf-viewer' Using host libthread_db library "/lib/libthread_db.so.1". [Thread debugging using libthread_db enabled] [New Thread 16384 (LWP 32022)] 0x40ff3018 in waitpid () from /lib/libpthread.so.0
+ Trace 45828
Thread 1 (Thread 16384 (LWP 32022))
Which version of gpdf is this?
It has to be the latest GPDF available to Gentoo, which I believe is 0.131. I should emphasize that after more testing, it appears that this crash only occurs with the PDFs mentioned in this report, not ones from other sites. (Strange) As a side note, I tried to report the version in which this error happens, but GPDF's bugzilla page does not have itself scripted to report the product versions; I suggest adding this feature.
Reopen because the information was supplied, and I can reproduce a crash.
I start working on this. The thumbs embed in the pdf seems to have a strange format. At least one filter have a type named "name". I not yet sure if the file have been misgenerated or if the ThumbColormap object have a parse bug. At least it doesn't crash anymore ... I take it...
Remi, can you please check the patch I did for bug #139211? It fixed this bug as well. It seems the old code handled only Indexed color spaces, but not DeviceGray or DeviceRGB. *** This bug has been marked as a duplicate of 139211 ***
The bug was in colormap filter parsing (using only array objects instead of also accepting name objects when only one filter). Using Gfx colormap is ok (you removed the bug) but you also removed the optimisation I added for thumb colormap (most of the thumbs are created with a common colormap). Creating one colormap at each reference is really a big job ... (I never saw a thumb created with a none indexed CM. This is why I created my own simpler class). By removing it I'am afraid that the thumbs processing become still longer (and it is already a pain ;-)