Bug 449734

+++ This bug was initially created as a clone of Bug #448911 +++

Description of problem:
File conflict between ia64 (primary) and i386 (compatibility) version of the
package.

Version-Release number of selected component (if applicable):
pam-0.77-66.25

How reproducible:
100%

Steps to Reproduce:
1. Install the ia64 version of the package
2. Install ia32el and verify that it is running
3. Install the i386 version of the package
  
Actual results:
file /usr/share/doc/pam-0.77/pdf/pam.pdf from install of pam-0.77-66.25
conflicts with file from package pam-0.77-66.25

file /usr/share/doc/pam-0.77/pdf/pam_appl.pdf from install of pam-0.77-66.25
conflicts with file from package pam-0.77-66.25

file /usr/share/doc/pam-0.77/pdf/pam_modules.pdf from install of pam-0.77-66.25
conflicts with file from package pam-0.77-66.25

file /usr/share/doc/pam-0.77/ps/pam.ps from install of pam-0.77-66.25 conflicts
with file from package pam-0.77-66.25

file /usr/share/doc/pam-0.77/ps/pam_appl.ps from install of pam-0.77-66.25
conflicts with file from package pam-0.77-66.25

file /usr/share/doc/pam-0.77/ps/pam_modules.ps from install of pam-0.77-66.25
conflicts with file from package pam-0.77-66.25


Expected results:
no file conflicts

Additional info:
This is a regression against RHEL4-U6

-- Additional comment from atodorov on 2008-05-29 09:21 EST --
KATE result failing with latest compose:
http://hank.test.redhat.com/kate/auto.cgi?action=case&group=RHEL4-U7-re20080528.1-ia64&case=rhel4-ia64-compat-layer-AS-releng-RHEL4-U7-LACD-re20080519.-releng-english-nfs-GUI-_everything

KATE result passes with 4.6:
http://hank.test.redhat.com/kate/auto.cgi?action=case&group=RHEL4-U6-re20071018.0-ia64&case=rhel4-ia64-compat-layer-AS-releng-RHEL4-U6-LACD-re20071018.-releng-english-nfs-GUI-_everything

-- Additional comment from atodorov on 2008-05-29 09:22 EST --
requesting RC blocker

-- Additional comment from tmraz on 2008-05-29 10:11 EST --
This is not a regression in the pam package. These files were always different
between ia64 and i386 pam package. Perhaps something else changed between 4.6
and 4.7? rpm possibly?


-- Additional comment from atodorov on 2008-05-29 10:49 EST --
This is a regression in pam. For example the conflicting file above
/usr/share/doc/pam-0.77/ps/pam_modules.ps

is present in the ia64 and i386 version in RHEL4/U6 and doesn't conflict.

The diff between the same file in 4.7 is:

--- pam_modules.ps.ia64 2008-03-27 13:52:25.000000000 +0100
+++ pam_modules.ps.i386 2008-03-27 13:42:49.000000000 +0100
@@ -8,10 +8,10 @@
 %%EndComments
 %DVIPSWebPage: (www.radicaleye.com)
 %DVIPSCommandLine: dvips -R -q -t letter -o
-%+ /tmp/linuxdoc-dir-15919/sgmltmp.pam_modules-latex-15919.dir/pam_modules.ps
+%+ /tmp/linuxdoc-dir-2234/sgmltmp.pam_modules-latex-2234.dir/pam_modules.ps
 %+ pam_modules.dvi
 %DVIPSParameters: dpi=600, compressed
-%DVIPSSource:  TeX output 2008.03.27:0852
+%DVIPSSource:  TeX output 2008.03.27:0842
 %%BeginProcSet: texc.pro
 %!
 /TeXDict 300 dict def TeXDict begin/N{def}def/B{bind def}N/S{exch}N/X{S

clearly they don't differ in content.

-- Additional comment from tmraz on 2008-05-29 11:04 EST --
NO!
This is NOT a regression in PAM - look at the pam-0.77-66.23 (version from RHEL4/U6)
these files were always different between architectures.


-- Additional comment from pmatilai on 2008-05-29 12:44 EST --
The conflict might have always been there, it's just that in RHEL 4.7 rpm no
longer completely ignores them. Previously all sorts of checking was skipped for
large number of directories, such as docdirs, translations etc.

-- Additional comment from tmraz on 2008-05-30 03:14 EST --
Yes, that's what I expected. Unfortunately to solve this problem the pam package
would have to be changed to split out the documentation into a separate
subpackage. Would that be appropriate for RHEL-4.7 package? When the public beta
is already out so possible problems with this split will not be found by public
testing?

It was really unfortunate to introduce this kind of change into RHEL-4.7 rpm
without doing thorough check of shipped packages for conflicts in advance. We
could fix them appropriately and with appropriate amount of testing, not this
late in the RHEL-4.7 development cycle.

If this should be fixed in RHEL-4.8, it should be fixed in RHEL-5.3 as well so
we do not regress in this regard.