Build Error (Job 3064): mythtv-0_22-0_1_svn_r20232_fc11 on fedora-development-rpmfusion_free

[Jarod Wilson]

On Sunday 22 March 2009 19:25:31 Dominik 'Rathann' Mierzejewski wrote:
> On Sunday, 22 March 2009 at 21:47, Jarod Wilson wrote:
> > rpmfusion-buildsys at lists.rpmfusion.org wrote:
> > >Job failed on arch x86_64
> > >
> > >
> > >         Build logs may be found at
> > >        
> > > http://buildsys.rpmfusion.org/logs/fedora-development-rpmfusion_free/30
> > >64-mythtv-0.22-0.1.svn.r20232.fc11/
[...]
> > Looking for some clues here... A lock f11 x86_64 mock build works just
> > fine, as do builds on an actual f11 x86_64 machine (as well as f10
> > x86_64 machines). Why on earth is it failing in the buildsystem?
>
> Could be a different release of gcc rpm (check root.log).

Same gcc. Heck, so far as I can see, same *everything*. The only difference I 
see in the packages installed in root.log is that ccache is present on this 
end, wasn't on that end. Guess up next is a mock build attempt w/ccache 
disabled...

> Still, mythtv shouldn't be bundling libavutil,

Yes, so I've heard, many times. :) Not likely to change anytime soon though.

> which defines some system functions
> if they're not available (based on detection by FFmpeg's configure
> script.

Yep, bits of math.h. But those bits are most definitely defined on a rawhide 
system, and by a header most definitely installed at build time, so its quite 
baffling.

> If mythtv's snapshot is badly integrated or buggy then this
> might be a side-effect.

Given that it builds correctly everywhere else, I'm more inclined to think its 
a quirk of the build system, but I'll withhold judgment until I've tried a 
mock build w/o ccache involved. Of course, this was a clean chroot, never 
before used, and mythtv hadn't been compiled on this system ever before, so 
I'm not quite sure how ccache would have magically made things work... Bleah.

--jarod