On 17 October 2016 at 19:07, Marek Olšák <mar...@gmail.com> wrote:
> On Mon, Oct 17, 2016 at 7:10 PM, Emil Velikov <emil.l.veli...@gmail.com>
>> On 17 October 2016 at 17:47, Marek Olšák <mar...@gmail.com> wrote:
>>> From: Deepak Sharma <deepak.sha...@amd.com>
>>> This allows building radeonsi driver for surfaceless platform.
>>> Signed-off-by: Deepak Sharma <deepak.sha...@amd.com>
>>> Signed-off-by: Marek Olšák <marek.ol...@amd.com>
>> As I "predicted" from the beginning adding require_egl_drm() is a hack
>> which will inspire even more hacks.
>> For example 98cebc913cd7ad07ce9699b5f63d72fe3969c851 and now this one.
>> Please _drop_ these all together.
>> I realise what you're trying to be extra careful on behalf of the
>> user, but this is not the way to do it. Let's drop the "interesting"
>> enable-gbm auto-detection enabling it by default or toggling it on
>> supported platforms - !cygwin/windows and add the drm egl-platform if
>> gbm is enabled. This way your worries will be resolved, plus things
>> will be shorter and easier to untangle :-)
> While I understand where your criticism is coming from, this is not a
> hack to me. It's really just an improvement to already-good code. ;)
As a person who's been staring at this and similar pieces for far too
long, I can guarantee you it _is_ a hack. If you want to call it any
other way, so be it.
The rule is that you cannot enforce runtime specifics at compile time.
Similar to the LLVM topic.
If you want to go forward proceed with the earlier suggestion or
propose an alternative one, please ?
mesa-dev mailing list