[poppler] pdftoppm hang

[Albert Astals Cid]

El Dimarts, 24 d'abril de 2012, a les 10:11:02, Thomas Freitag va escriure:
> Am 23.04.2012 22:34, schrieb Albert Astals Cid:
> > El Diumenge, 22 d'abril de 2012, a les 15:56:51, Thomas Freitag va 
escriure:
> >> Am 11.04.2012 15:49, schrieb Thomas Freitag:
> >>> Am 11.04.2012 15:35, schrieb Ralph:
> >>>> Hi Folks,
> >>>> 
> >>>> I have a PDF file that's causing pdftoppm to hang.  Looking at an
> >>>> strace shows that it's doing *something* but it's taking much longer
> >>>> than expected (at least over a few minutes).  This is present on
> >>>> 0.18.3 and 0.19.2.
> >>>> 
> >>>> I would attach the file to a bug post, but I'd rather email it out
> >>>> directly as it's a customers PDF and not mine.  Who should I send
> >>>> this bugger to?
> >> 
> >> In the meantime I was able to solve it or at least make it better.
> >> Because my code base is completely different in the meantime, I attach
> >> here an inofficial complete patch for all my code changes. I regtested
> >> it, and there is one (only!) last unacceptable change in it
> >> (bug-poppler24733.pdf on page 3). I also made several valgrind checks on
> >> it, so there are probably no memory leaks in it.
> >> Because I've to solve this last regression first, and because Albert's
> >> vacation in May, we'll probably could not commit any of the five changes
> >> before 0.20.0,
> > 
> > I'm actually not on vacation, just working abroad ;-)
> > 
> >> but perhaps someone wants use the attached patch to see
> >> the enhancements.
> >> 
> >> What does it contains?
> >> 1. The DeviceN implementation in the splash backend. With it there also
> >> no more known problems with overprint.
> > 
> > Great :-)
> > 
> >> 2. A revision of the implementation of blend modes in the CMYK (and
> >> DeviceN) branch in splash.
> > 
> > Great :-)
> > 
> >> 3. A detection of unnecessary tranparency groups according note 5 of the
> >> PDF spec, which dramatically can speed up the rendering of several PDF,
> >> s. Bug 13487 (only splash!)
> > 
> > Great :-)
> > 
> >> 4. An implementation of knockout transparency groups in splash, Bug 12185
> > 
> > Great :-)
> > 
> >> 5. An optimization for image masks in pattern colorspace in splash, Bug
> >> 48964
> > 
> > Great :-)
> 
> After I was able to fix also the last regression I encountered, I
> started now with this one, even if it is the newest. The patch is
> attached to the call and is a little bit redesigned to make the
> implementation cleaner.

Patch looks good (i.e. has no regressions), but i'd like to have access to 
some file that proves this is faster than the old code before commiting it.

Cheers,
  Albert

> 
> Cheers,
> Thomas
> 
> >> Albert, I think the best way would be to deliver this in 5 patches, but
> >> because the changes at least share one source file, I must do it step by
> >> step. What order would You prefer?
> > 
> > I don't really have any order of preference, the one you prefer will suit
> > me.
> > 
> > Good work ;-)
> > 
> > Cheers,
> > 
> >    Albert
> >> 
> >> Cheers,
> >> Thomas
> >> 
> >>> If You don't care, send it to me in private. Then I'll have a look at
> >>> it probably next weekend.
> >>> 
> >>> Thomas
> >>> 
> >>>> Thanks for all the help,
> >>> 
> >>> _______________________________________________
> >>> poppler mailing list
> >>> poppler at lists.freedesktop.org
> >>> http://lists.freedesktop.org/mailman/listinfo/poppler
> >>> 
> >>> .
> > 
> > _______________________________________________
> > poppler mailing list
> > poppler at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/poppler
> > 
> > .
> 
> _______________________________________________
> poppler mailing list
> poppler at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/poppler