[Mesa-dev] [PATCH] gallium: build virgl by default.

[Emil Velikov]

On 23 November 2015 at 07:57, Albert Freeman <albertwdfreeman at gmail.com> wrote:
> On 23 November 2015 at 03:17, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
>> On Sun, Nov 22, 2015 at 9:28 PM, Dave Airlie <airlied at gmail.com> wrote:
>>> From: Dave Airlie <airlied at redhat.com>
>>>
>>> No reason not to build this, it has no external
>>> requirements at the moment.
>>
>> Not that I object, but what's the policy around these defaults? Should
>> we also add nouveau? freedreno? vc4?
>>
Yes please.

For those wondering about radeonsi, i915g or ilo. The first one has
llvm dependency which we cannot enforce onto the unaware user. With
the latter two - while I cannot comment on their quality/completeness,
they see almost no action (dev. wise) plus having them will bring
serious classic vs gallium confusion.

>> Of course my personal preference would be to just error the build if
>> no drivers are specified... but I guess that won't work for some
>> people?
Ideal case will be to 1) select everything regardless of dependencies
or 2) select the ones that have their dependencies resolved. I'm
leaning towards the latter, although that'll need a lot churn.

> I do agree with this, it will cause some of those people to look
> through all the gallium drivers to see which ones they need (maybe
> that will result in them using more...), reduce compilation time and
> decrease used storage space. It should also prevent people from not
> including the drivers they need (only for gallium drivers). Of course
> it will break a few peoples builds (though they will also reap the
> benefits), but almost everyone uses the flag.
>
In practise most (99%?) people building mesa provide their own drivers
list. The default list covers the remaining people, which have little
to no idea what they want/need. Thus building everything * (see note
above) is the reasonable thing to do.

Once people get the feel - "I want vc4/nouveau/etc", they can tweak
the list and get compilation time/size savings.

> The only other fair alternative is to build all the drivers, which has
> a negative impact on all those points (excluding broken builds) +
> dependency issues (other broken builds (less though: some people would
> have the required dependencies)).
>>
>>>
>>> Signed-off-by: Dave Airlie <airlied at redhat.com>

Dave feel free to add:

Reviewed-by: Emil Velikov <emil.l.velikov at gmail.com>

-Emil