[Nouveau] 3.19.0-rc1 nouvea build failure on GeForce GT 610 only

[Paul Bolle]

On Mon, 2014-12-22 at 14:37 +0100, Paul Bolle wrote:
> On Mon, 2014-12-22 at 13:01 +0000, Sid Boyce wrote:
> >    CHK     kernel/config_data.h
> >    CC [M]  drivers/gpu/drm/nouveau/core/engine/dmaobj/nvd0.o
> > drivers/gpu/drm/nouveau/core/engine/dmaobj/nvd0.c: In function 
> > ‘nvd0_dmaobj_bind’:
> > drivers/gpu/drm/nouveau/core/engine/dmaobj/nvd0.c:54:8: error: 
> > ‘GM204_DISP_CORE_CHANNEL_DMA’ undeclared (first use in this function)
> >     case GM204_DISP_CORE_CHANNEL_DMA:
> >          ^
> > drivers/gpu/drm/nouveau/core/engine/dmaobj/nvd0.c:54:8: note: each 
> > undeclared identifier is reported only once for each function it appears in
> > scripts/Makefile.build:257: recipe for target 
> > 'drivers/gpu/drm/nouveau/core/engine/dmaobj/nvd0.o' failed
> > make[4]: *** [drivers/gpu/drm/nouveau/core/engine/dmaobj/nvd0.o] Error 1
> > scripts/Makefile.build:402: recipe for target 'drivers/gpu/drm/nouveau' 
> > failed
> > make[3]: *** [drivers/gpu/drm/nouveau] Error 2
> > scripts/Makefile.build:402: recipe for target 'drivers/gpu/drm' failed
> > make[2]: *** [drivers/gpu/drm] Error 2
> > scripts/Makefile.build:402: recipe for target 'drivers/gpu' failed
> > make[1]: *** [drivers/gpu] Error 2
> > Makefile:938: recipe for target 'drivers' failed
> > make: *** [drivers] Error 2
> > 
> > 01:00.0 VGA compatible controller: NVIDIA Corporation GF119 [GeForce GT 
> > 610] (rev a1)
> 
> Why would that controller influence the build?
> 
> > No problems on 2 other boxes with GeForce 8600 GT] (rev a1)
> 
> I ran into this error twice, while building Fedora 21 kernel rpms for
> v3.19-rc1 on my outdated ThinkPad X41 (which uses i915, for what it's
> worth). The error made very little sense to me. Especially since it
> didn't happen when building in the local git repository.
> 
> So I basically, well, scrubbed the rpmbuild environment. And then the
> the third attempt to build kernels rpm succeeded! I assumed a hardware
> failure (see outdated above) and made a mental note to investigate that
> one of these days.
> 
> But your report makes me think it's not my hardware. Could there be some
> _random_ failure with this driver in the build? Very odd...

A theory based on Mark's message in
https://lkml.org/lkml/2014/12/8/602 :

My two failed builds somehow used previous version of class.h, that
didn't yet have commit 1f89b4756fb8 ("drm/gm204/disp: initial support")
applied. It's that commit which added the define for
GM204_DISP_CORE_CHANNEL_DMA.

Is that even possible? Perhaps rpmbuild has a surprising way to handle
symlinks? All very puzzling...


Paul Bolle