> -----Original Message-----
> From: Burton, Ross [mailto:ross.bur...@intel.com]
> Sent: Friday, October 12, 2012 2:49 AM
> To: Kamble, Nitin A
> Cc: Zanussi, Tom; Hart, Darren; yocto@yoctoproject.org
> Subject: Re: [yocto] [PATCHv4 6/8] mesa-dri.bbappend: avoid conflict with
> emgd-driver-bin
> 
> On 12 October 2012 00:31,  <nitin.a.kam...@intel.com> wrote:
> > Extend the mesa-dri recipe from oecore to avoid conflict with files
> > generated by emgd-driver-bin recipe.
> 
> The commit message should also say what the high level changes were, in
> this case
> 

The information about high level change is in the next lines in the comment. As 
seen in the lines bellow here.

> > +# To avoid conflict disable egl, gles1, gles2 from meta-dri if the
> > +BSP image # is bundling the emgd driver.
> 
> I'm not entirely keen on this level of deep hackery, it looks very
> fragile. 

What seems like the hackery here, is simple change to EXTRA_OECONF variable of 
the mesa-dri recipe. But the way original mesa-dri recipe is written, there is 
no simpler bitbake way to change EXTRA_OECONF.

>  Is there any difference in the headers installed by emgd
> and mesa?  

Yes, there are differences, the EMGD recipe is providing GL implementation at 
ABI level which has not going in the open source version of mesa-dri.

> An alternative would be to consider mesa a more canonical
> source of GL headers for the build, and change EMGD to not install
> anything to staging.   This would mean that emgd wouldn't build-time
> provide any virtual/libgl packages, just runtime provides so that the image
> can be built using EMGD and Mesa's libgl.
> 

There is a reason EMGD recipe is providing these GL libraries in binary form 
and not in the source form, I think there is EMGD specific proprietary code in 
these libraries. I can check with EMGD guys to confirm this. 

> FYI, In ross/xorg I've started re-arranging the packaging of mesa, so that the
> packages are called libgl-mesa and RPROVIDE libgl.  I plan to make the same
> changes to EMGD once I have the shlib resolution sorted.
> 

Looks like this kind of packaging change would not change functionality of both 
recipes. 

Cheers,
Nitin


> Ross
_______________________________________________
yocto mailing list
yocto@yoctoproject.org
https://lists.yoctoproject.org/listinfo/yocto

Reply via email to