Description of problem:
During generating documentation of libisofs-devel-0.6.32 using doxygen there is a multilib conflict caused by different PNG images generated on i386 and x86_64. The problem is not caused by different timestamp, the images are really different (arrows are shifted by a couple of pixels). The mentioned multilib conflict is reported in bug #658034. I think PNG files should be the same regardless of arch.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. build libisofs-0.6.32 on i386 and x86_64
2. compare image /usr/share/doc/libisofs-devel-0.6.32/html/structiso__stream__coll__graph.png in both rpms
PNG images differ.
PNG images are same.
This request was evaluated by Red Hat Product Management for
inclusion in the current release of Red Hat Enterprise Linux.
Because the affected component is not scheduled to be updated
in the current release, Red Hat is unfortunately unable to
address this request at this time. Red Hat invites you to
ask your support representative to propose this request, if
appropriate and relevant, in the next release of Red Hat
Enterprise Linux. If you would like it considered as an
exception in the current release, please ask your support
it's a bug in graphviz, doxygen calls graphviz tools to create the png files. A workaround for libisofs is to set HAVE_DOT= no in doxygen.conf.in to not use dot from graphviz.
Analysis of this problem written in Bug 827927 comment 5.
Technical note added. If any revisions are required, please edit the "Technical Notes" field
accordingly. All revisions will be proofread by the Engineering Content Services team.
Previously dot could generate different images on i386 and x86_64 architectures, which could lead to multilib conflicts for packages that use graphviz during its build process. The problem was caused by i386 FPU. It's registers are wider than C double type and if compiled with -O2 more arithmetic is done in FPU registers. The error increases which finally could lead to round/trunc error and shift in output images. Now the code is compiled with -ffloat-store compiler flag on i386 and intermediate results are not stored in FPU registers. Also the -ffast-math compiler flag was removed, because the performance benefit is very low, but it could cause a lot of hard to debug problems.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.