RHEL bug, how to proceed?

[Panu Matilainen]

On 04/24/2013 05:49 PM, Richard W.M. Jones wrote:
> On Wed, Apr 24, 2013 at 04:35:24PM +0200, Simone Caronni wrote:
>> Hello,
>>
>> I have a package that needs procps. While building it in EPEL, i stepped
>> onto a packaging problem [1] in the RHEL 6 package.
>>
>> I've opened the bug, but nobody is looking at it and (my guess) it will be
>> probably fixed for RHEL 6.5. Note that the bug has been tagged as "EasyFix"-
>>
>> The workaround for the problem would be as follows:
>>
>> %if 0%{?rhel} == 6
>> BuildRequires:          procps
>> %else
>> BuildRequires:          procps-devel
>> %endif
>>
>> I suppose there's no "provenpackager" group in RHEL, so what should I do in
>> this case? Build the package with a workaround in place until the bug is
>> fixed or wait on building the package at all?
>>
>> Thanks,
>> --Simone
>>
>> [1] https://bugzilla.redhat.com/show_bug.cgi?id=950748
>
> It's definitely a bug.  The spec file has:
>
>    %files
>    %attr(755,root,root) /%{_lib}/*
>    [...]
>    %files devel
>    /%{_lib}/libproc.so
>
> which means that libproc.so is included in both RPMs.  Although this
> doesn't stop installation for me -- RPM notices that both files are
> the same and allows it.  I don't know why mock disallows this.

[pmatilai at turre tmp]$ rpm -qplv --nosignature 
procps-devel-3.2.8-25.el6.x86_64.rpm procps-3.2.8-25.el6.x86_64.rpm 
|grep proc.so
lrwxrwxrwx    1 root    root                       16 Nov 15 15:54 
/lib64/libproc.so -> libproc-3.2.8.so
lrwxr-xr-x    1 root    root                       16 Nov 15 15:54 
/lib64/libproc.so -> libproc-3.2.8.so
[pmatilai at turre tmp]$

The mode differs which is a conflict in rpm >= 4.10, older ones silently 
slip it through.

There's an apparent rpm bug or two here as well: symlink permissions are 
not meaningful so they shouldn't conflict, and the latter "lrwxr-xr-x" 
mode is not what is expected for symlinks.

	- Panu -