rpmbuild of rawhide kernel fails.

[Josh Boyer]

On Tue, Feb 17, 2015 at 2:50 PM, Paul Bolle <pebolle at tiscali.nl> wrote:
> On Tue, 2015-02-17 at 11:54 -0500, Steve Dickson wrote:
>> I'm trying to build a rawhide kernel
>> (kernel-3.20.0-0.rc0.git7.1.fc23.src.rpm)
>>
>> with rpmbuild -bb --target x86_64 kernel.spec
>> and I'm getting the following error
>>
>> + '[' '!' -f /data/src/rawhide/build/SOURCES/patch-3.19-git7.xz ']'
>> + case "$patch" in
>> + unxz
>> + patch -p1 -F1 -s
>> symbolic link target '../../../../../arch/powerpc/lib/memcmp_64.S' is invalid
>> error: Bad exit status from /var/tmp/rpm-tmp.EsUbKg (%prep)
>>
>>
>> RPM build errors:
>>     Bad exit status from /var/tmp/rpm-tmp.EsUbKg (%prep)
>>
>> All the output is at http://ur1.ca/jqwao
>>
>> Any ideas why this is happening?
>
> This reminds me of a similar rpmbuild failure I ran into two months ago,
> also involving symlinks: https://lkml.org/lkml/2014/12/22/177 . My
> current theory is that there's something odd going on at the
> intersection of rpmbuild, cp -l, and/or patch.
>
> But, of course, replaying part of your rpmbuild output didn't trigger
> this problem on a f20 system:
>     $ cp -al vanilla-3.19 vanilla-3.19-git7
>     $ cd vanilla-3.19-git7/
>     $ cat [...]/patch-3.19-git7.xz | unxz | patch -p1 -F1 -s
>     $ file tools/testing/selftests/powerpc/stringloops/memcmp_64.S
>     tools/testing/selftests/powerpc/stringloops/memcmp_64.S: symbolic link to `../../../../../arch/powerpc/lib/memcmp_64.S'
>
> Still all very puzzling...

In rawhide, patch-2.7.3 stopped supporting relative symlinks.  Then
patch-2.7.4 went back to supporting symlinks.  So make sure you have a
patch package installed that isn't 2.7.3.

josh