Red Hat Bugzilla – Bug 225069
Font objects in pdf files get cut at conversion to other formats
Last modified: 2008-01-14 07:23:02 EST
Description of problem:
Certain pdf files generated by music typesetting utilities like Finale and
Sibelius get big size objects belonging to embedded fonts cut off during
conversion to other formats.
That happens for example when processing pdf files with gimp (pdf -> png) and
printing out with cups
Version-Release number of selected component (if applicable):
The errors occurs consistently
Steps to Reproduce:
1. Open pdf file through gimp file open dialogue
2. The cut-off errors are immediately visible
See attached pdf test case file La_Fille.pdf and page 1 converted to png
See attached file La_Fille-preview-AFPL-Ghostscript-8.54.png
As demonstrated in the attached file,
La_Fille-preview-AFPL-Ghostscript-8.54.png I am able to circumvent the problem
in gimp context by setting the PATH and GS_LIB environments to point to
AFPL-Ghostscript-8.54 built from source distribution and installed in
a /usr/local subdirectory
Created attachment 146765 [details]
Test case for demonstrating reported bug
Created attachment 146766 [details]
conversion of test pdf through ESP Ghostscript 815.02
Created attachment 146767 [details]
conversion of test pdf through AFPL Ghostscript 8.54
I think I need to be a little more specific as to the observed conversion
The test case is sheet music for piano, typical for the pdf scores I prepare
daily for publishing in the non-commercial classical sheet music web site,
http://icking-music-archive.org/ which I'm editing.
Near the left border of each staff (5 parallele lines) there are so-called
clefs: a G-clef on the upper of 2 staves kept together by curled braces and an
F-cleft on the lower staff. If you compare the clef shapes as found in the pdf
file with those found in the png file converted through the ESP Ghostscript
815.02 coming with FC5, then you'll notice that the clef symbols are cut off
above, below and to the right, but not to the left. Contrarily in the png file
convertd through AFPL Ghostscript 8.54 the G- and F-clefs are rendered
My test pdf file has been typeset with the commercial, high-end music
typesetting utitily 'Finale 2005' using the type-1 music fonts
ZPCJVL+EngraverTextT TrueType yes yes no 8 0
VFYNOX+EngraverFontSet TrueType yes yes no 6 0
XUCZVZ+EngraverFontSet TrueType yes yes yes 10 0
I have observed similar conversion error with pdf scores genereated from other
music typesetting programs based on rendering music symbols through fonts.
While I, as indicated, am able to circumvent the ghostscript error in gimp
context, I've until now not been able to figure out how to solve the problem
in cups context.
I recently upgraded from FC5 where I did not encounter that problem
Thanks for the detailed test case.
Can you confirm the last release of Fedora that worked correctly for you? This
bug is filed against FC5, and the specific RPMs are from FC5, but your last
comment about upgrading has me confused.
(In reply to comment #5)
> Thanks for the detailed test case.
> Can you confirm the last release of Fedora that worked correctly for you? This
> bug is filed against FC5, and the specific RPMs are from FC5, but your last
> comment about upgrading has me confused.
Sorry for confusing you. What I should have written is that I recently upgraded
from FC 4 to FC 5 (not from FC 5 to FC 6). Unfortunately It is not possible for
me to check the problem in an FC 6 environment.
A few days ago I coincidentally became aware of the cups configuration file
/etc/foomatic/filter.conf. In that file I've explicitly set the variable gspath
to point to the gs binary as installed with AFPL-Ghostscript-8.54. As I had
hoped the reported problem with cutting off font shapes has been amended. But
now cups printing can't cope any more with documents in landscape format.
Nothing gets printed. This is a minor drawback, though. But I've got my
suspicion with ESP Ghostscript 815.02 confirmed.
Created attachment 150314 [details]
New test case, input file for test
A WIMA contribution which I received today reveals additional aspects of the
reported gs problems. Here follows the pdf score I received.
Created attachment 150316 [details]
result of applying gs-8.15 pdf2ps
Notice how full page width areas are cut out by pdf2ps processing.
Created attachment 150317 [details]
result of applying gs-8.54 pdf2ps
this file shows what the result of pdf2ps should be.
Works correctly in Fedora 8.