Re: review of OpenGL blacklist work

On Fri, Jun 5, 2015 at 2:16 PM, Michael Stahl <mstahl@redhat.com> wrote:

> On 05.06.2015 13:59, Thorsten Behrens wrote:
> > Markus Mohrhard wrote:
> > > I have implemented something similar now in the
> feature/opengl-preparation
> > > branch for windows as preparation for the glyphy work (which is
> expected to
> > > uncover many driver bugs). I'd appreciated if someone would have a look
> at
> > > it and comment on the general idea (do we want to use that concept, is
> > > there something that I missed, ...).
> > Looks good in general, just of course the devil is in the details -
> >
> > Ideally, one would either have that within the established libreoffice
> > xml configuration system (such that system integrators can override
> > it, lock it down partially, install config-only extensions that tweak
> > it etc etc) - or have an optional config layer that overrides whatever
> > special downloaded blacklist there is. See
> > canvas/source/directx/dx_config.cxx for an existing solution.
>
> iirc the OpenCL stuff also has some sort of blacklist nowadays, wonder
> if all of these should use the same sort of mechanism - that would make
> it easier to admin.

Sadly the OpenCL stuff uses a manual blacklist that has to be maintained by
the user. I don't like that solution as at least 99% of the user don't
understand how to handle that. For OpenCL we also have the option of
running a short test whenever the driver/device changes. That was my
original idea for the OpenCL blacklist when it was discussed.

Of course we should make sure that if we want we can use the same mechanism
for other features as well.

Markus