F21: building kernels from srpms is broken

[Steve Dickson]

On 08/26/2014 01:54 PM, Josh Boyer wrote:
>> Yes.. Your advice was to use 'fedpkg local' which does indeed work but
>> > it also builds a bunch of rpms I don't need or use, like debug kernels.
> Which you failed to tell us yesterday.  Come on Steve.
Sorry about that... I really didn't think of it until the build
took 3 times as long as it usually did... 

> 
>> > When I rpmbuild from a srpm I use the following flags
>> >    --without extra --without tools --without debug --without perf --without headers
> I'm guessing it's the --without extra that's breaking things.  gfs2.ko
> is typically shoved in the modules-extra package.  You aren't actually
> saving any build time by passing that, since everything in there is
> still getting built.  It's just winding up in some other package.  We
> should probably just remove that config switch because it really
> doesn't make sense.
Yes.. have things break early on would be better... IMHO...

> 
> The --without headers isn't going to save you much time either.
Ack..  

> 
>> > which greatly decrease the compiling time... from 70mins to 30mins
>> > depending on the machine...
>> >
>> > Now if I can set those flags in the fedpkg local build... Then
>> > I'm good to go...
>> >
>>> >>
>>> >> I guess we can start over.  Can you tell us exactly what you
>>> >> downloaded from where, and how you're invoking the build?  This
>>> >> obviously builds in koji, and it's built multiple times locally as
>>> >> well.  It's rather confusing that you're having troubles, so we need
>>> >> as much info on what you're doing as possible so we can try and
>>> >> recreate.
>> > All I'm trying to do is build from a source rpm so I can set those flags.
>> > Here is what I'm doing... If there is a better way please let me know...
>> >
>> > fedpkg clone kernel
>> > cd kernel
>> > fedpkg srpm
>> > rpm -ihv kernel*.rpm
>> > cd /home/src/SPEC
>> > rpmbuild -bb --without extra --without tools --without debug --without perf --without headers --target x86_64 kernel.spec
>> >
>> > I've been build kernels this way for quite a while... but again,
>> > if there is a better way I'm more than willing to try...
> Those steps are fine, with the caveat about --without extras from
> above.  if you had told us all of this yesterday, you would have had
> an answer yesterday too.
Well I guess things have worked so well... for so long... I was
not sure what was or was not relative... But it all does make
sense now... 

Thanks!

steved.