Red Hat Bugzilla – Bug 827927
Doxygen causes slightly different images on i686 and x86_64
Last modified: 2013-08-01 13:16:47 EDT
Description of problem:
The x86_64 -devel package has files which conflict with those in the i686
Version-Release number of selected component (if applicable):
Everytime, attempt to install both packages on an x86_64 system.
Doxygen causes slightly different images on i686 and x86_64.
Doxygen should cause same images on i686 and x86_64.
File conflict for /usr/share/doc/popt-devel-1.13/html/findme_8c__incl.png in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/findme_8h__dep__incl.png in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/poptint_8h.html in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/poptint_8h__dep__incl.map in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/poptint_8h__dep__incl.png in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/poptparse_8c__incl.png in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/structpoptContext__s__coll__graph.png in popt-devel-1.13-10.fc17
File conflict for /usr/share/doc/popt-devel-1.13/html/system_8h__incl.png in popt-devel-1.13-10.fc17
Created attachment 588919 [details]
Created attachment 588920 [details]
The changes are like:
-<area shape="rect" id="node3" href="$popt_8c.html" title="popt.c" alt="" coords="4,80,61,107"/>
+<area shape="rect" id="node3" href="$popt_8c.html" title="popt.c" alt="" coords="5,80,63,107"/>
Created attachment 591498 [details]
This is fix for the *.html difference problem - remove -ffast-math.
Verified the problem happens with kept -ffast-math and the problem disappears after removing -ffast-math.
Unfortunately there still remains a problem with different PNGs:
file /usr/share/doc/popt-devel-1.13/html/nav_g.png conflicts between attempted installs of popt-devel-1.13-10.fc18.i686 and popt-devel-1.13-10.fc18.x86_64
file /usr/share/doc/popt-devel-1.13/html/structpoptContext__s__coll__graph.png conflicts between attempted installs of popt-devel-1.13-10.fc18.i686 and popt-devel-1.13-10.fc18.x86_64
Neither libpng nor netpbm uses -ffast-math, not sure why that can happen (and what is the real producer of these .png files).
Created attachment 600245 [details]
Reproducer for problem 2
Currently it seems there are (at least) 3 different problems:
1) The --ffast-math is ugly and needs to be removed. It will fix some wrong files (e.g. some PNG files in popt-devel).
2) On i686 if compiled with -O2, the compiler optimizes the code to do more arithmetic in FPU registers (which are wider than C float/double types), so there are less load/store that normally truncate the wider floats. As the error accumulates in FPU registers it finally causes several trunc/round errors. We can enforce the gcc to issue explicit load/store on problematic parts of code, but the beahviour of gcc optimizer could change with new versions and such hacks could became non-effective. Reproducer for this problem is attached (dot.dot). Other, probably more safe solutions for this problem are -O0 or --ffloat-store. The latter seems to have less performance penalty (because there are only some more load/store). I created very simple benchmark (root.dot from graphviz gallery, T500 computer, 32 bit, avg and stddev for 10 runs, stddev are written in braces, less numbers are better):
-O2 -ffloat-store -O0
4.305 (0.018) s 4.354 (0.040) s 4.696 (0.045) s
Also SSE instructions on 64 bit are faster in all cases, but that's another story (same test, same machine, 64 bit):
3.732 (0.021) s
3) It seems there is problem with pango rendering on F17. Currently I am not sure whether it is problem in pango itself or problem in pango support code in graphviz (will investigate). It is reproducible on structpoptContext__s__coll__graph.png from popt (I will attach the dot source for it later). It seems the source image differs in one pixel, i.e. the bmp check:
$ dot.32 structpoptContext__s__coll__graph.dot -Tbmp -o popt32.bmp
$ md5sum pop32.bmp
$ dot.64 structpoptContext__s__coll__graph.dot -Tbmp -o popt64.bmp
$ md5sum pop64.bmp
$ diff -u <(hexdump ./popt32.bmp) <(hexdump ./popt64.bmp)
--- /dev/fd/63 2012-07-25 10:47:11.063453467 +0200
+++ /dev/fd/62 2012-07-25 10:47:11.063453467 +0200
@@ -3810,7 +3810,7 @@
003b0c0 f5fd fffa ffff ffff ffff ffff ffff ffff
003b0d0 ffff ffff ffff ffff ffff ffff ffff ffff
-003b260 feff fefd 52d5 dcaa b86f ddf7 ffee ffff
+003b260 feff fefd 53d5 dcaa b86f ddf7 ffee ffff
003b270 ffff ffff ffff ffff ffff ffff beef cddf
003b280 9a32 4dd4 faa7 f4e9 ffff ffff ffff ffff
003b290 ffff ffff ffff ffff ffff ffff ffff ffff
Also same with PNG:
$ dot.32 structpoptContext__s__coll__graph.dot -Tpng -o popt32.png
$ md5sum popt32.png
$ dot.64 structpoptContext__s__coll__graph.dot -Tpng -o popt64.png
$ md5sum popt64.png
But it seems to be OK with gd:
$ dot.32 structpoptContext__s__coll__graph.dot -Tpng:gd:gd -o popt32.png
$ dot.64 structpoptContext__s__coll__graph.dot -Tpng:gd:gd -o popt64.png
This third problem is also not reproducible on RHEL-6, which prompts me that it may be pango problem.
Created attachment 600292 [details]
Reproducer for problem 3
The 3) seems to be an regression in graphviz (pango plugin). It is not reproducible with graphviz-2.26.0.
Fixes for 1) and 2) committed into rawhide/f18.
This message is a reminder that Fedora 17 is nearing its end of life.
Approximately 4 (four) weeks from now Fedora will stop maintaining
and issuing updates for Fedora 17. It is Fedora's policy to close all
bug reports from releases that are no longer maintained. At that time
this bug will be closed as WONTFIX if it remains open with a Fedora
'version' of '17'.
Package Maintainer: If you wish for this bug to remain open because you
plan to fix it in a currently maintained version, simply change the 'version'
to a later Fedora version prior to Fedora 17's end of life.
Bug Reporter: Thank you for reporting this issue and we are sorry that
we may not be able to fix it before Fedora 17 is end of life. If you
would still like to see this bug fixed and are able to reproduce it
against a later version of Fedora, you are encouraged change the
'version' to a later Fedora version prior to Fedora 17's end of life.
Although we aim to fix as many bugs as possible during every release's
lifetime, sometimes those efforts are overtaken by events. Often a
more recent Fedora release includes newer upstream software that fixes
bugs or makes them obsolete.
Fedora 17 changed to end-of-life (EOL) status on 2013-07-30. Fedora 17 is
no longer maintained, which means that it will not receive any further
security or bug fix updates. As a result we are closing this bug.
If you can reproduce this bug against a currently maintained version of
Fedora please feel free to reopen this bug against that version.
Thank you for reporting this bug and we are sorry it could not be fixed.