Hello. I encountered some errors in the 2.4.19-rc3 CML rules.
It took a little while to track down, but I finally know what's
going on:
1) CONFIG_DRM_SIS needs to require CONFIG_FB_SIS_315.
Currently, you can select CONFIG_DRM_SIS without CONFIG_FB_SIS_315.
If you do that, you get undefined symbol errors for sis_malloc and
sis_free.
2) CONFIG_FB_SIS must be compiled into the kernel (i.e. NOT a module).
Currently, you can compile it as a module.
If you do that, you ALSO get undefined symbol errors for
sis_malloc and sis_free.
These requirements could be enforced with CML rules. Before I
submit the patch to do this, I'd like to know if that's the proper
fix! Would it be better to just make CONFIG_FB_SIS able to be built
as a module instead?
Thanks for any help.
- Scott
On Wed, 2002-07-24 at 11:22, Scott Bronson wrote:
> 1) CONFIG_DRM_SIS needs to require CONFIG_FB_SIS_315.
> Currently, you can select CONFIG_DRM_SIS without CONFIG_FB_SIS_315.
> If you do that, you get undefined symbol errors for sis_malloc and
> sis_free.
>
> 2) CONFIG_FB_SIS must be compiled into the kernel (i.e. NOT a module).
> Currently, you can compile it as a module.
> If you do that, you ALSO get undefined symbol errors for
> sis_malloc and sis_free.
>
> These requirements could be enforced with CML rules. Before I
> submit the patch to do this, I'd like to know if that's the proper
> fix! Would it be better to just make CONFIG_FB_SIS able to be built
> as a module instead?
For the rules - go for it. For the modular driver ping the sisfb
maintainer first and check what is in the pipeline.
Alan
On Wed, 2002-07-24 at 11:22, Scott Bronson wrote:
> > 1) CONFIG_DRM_SIS needs to require CONFIG_FB_SIS_315.
> > Currently, you can select CONFIG_DRM_SIS without
> > CONFIG_FB_SIS_315. If you do that, you get undefined symbol
> > errors for sis_malloc and sis_free.
Not true. DRM_SIS does NOT require FB_SIS_315. It requires either
FB_SIS_300 *or* FB_SIS_315. (But since DRI is not supported on the 315
series, that whole issue does not make much sense...)
> 2) CONFIG_FB_SIS must be compiled into the kernel (i.e. NOT a
> module).
Not true either. If you compile DRM as a module as well, you don't get
any unresolved symbols. So it's either 1) both into the kernel, 2)
both as modules or 3) sisfb into the kernel and DRM as a module.
> > Currently, you can compile it as a module.
> > If you do that, you ALSO get undefined symbol errors for
> > sis_malloc and sis_free.
>
> These requirements could be enforced with CML rules. Before I
> submit the patch to do this, I'd like to know if that's the proper
> fix! Would it be better to just make CONFIG_FB_SIS able to be built
> as a module instead?
> For the rules - go for it.
No, please don't.
> For the modular driver ping the sisfb
> maintainer first and check what is in the pipeline.
Erm, that's me I think (apart from the fbdev api stuff)... Well, not
much for now as I am on vacation.
Thomas
--
Thomas Winischhofer
Vienna/Austria
mailto:[email protected] http://www.winischhofer.net/