[PATCH v3] winex11.drv: Make the GL context current even if no drawables were given.

[Matteo Bruni]

2016-02-26 10:42 GMT+01:00 Charles Davis <cdavis5x at gmail.com>:
>
>
> On Thu, Feb 25, 2016 at 7:29 AM, Matteo Bruni <matteo.mystral at gmail.com>
> wrote:
>>
>> 2016-02-24 5:12 GMT+01:00 Charles Davis <cdavis5x at gmail.com>:
>> > Signed-off-by: Charles Davis <cdavis5x at gmail.com>
>> > ---
>> > Try 3: Fix an assignment I missed. (Thanks, Ken.)
>> >
>> >  dlls/opengl32/tests/opengl.c |  2 +-
>> >  dlls/winex11.drv/opengl.c    | 69
>> > +++++++++++++++++++++++++-------------------
>> >  2 files changed, 41 insertions(+), 30 deletions(-)
>> >
>> > diff --git a/dlls/opengl32/tests/opengl.c b/dlls/opengl32/tests/opengl.c
>> > index e5e1507..c058c19 100644
>> > --- a/dlls/opengl32/tests/opengl.c
>> > +++ b/dlls/opengl32/tests/opengl.c
>> > @@ -975,7 +975,7 @@ static void test_opengl3(HDC hdc)
>> >           * expect drivers to ever offer it.
>> >           */
>> >          res = wglMakeCurrent(0, gl3Ctx);
>> > -        ok(res == FALSE, "Wow, OpenGL 3.0 windowless rendering passed
>> > while it was expected not to!\n");
>> > +        ok(res || broken(!res), "OpenGL 3.0 windowless rendering
>> > failed!\n");
>>
>> That seems suspicious...
>> AFAICS the comment right above this hunk is still valid (it is for
>> current Nvidia drivers on Win7 at least), which probably means that
>> this patch is not a good idea.
>
> Yeah, I was afraid of that.
>
> Truth be told, I don't need this for an application.
> Besides the fact that OpenGL 3.0+ is supposed to allow this,
> I added it because I thought it might be useful for wined3d (e.g. being able
> to create devices without an implicit swapchain, though now that I've looked
> into that possibility, making that change would be much harder than I thought). In any case, I'd
> imagine from your comment that you disagree. If you do, I'll
> retract this patch and revert the patch adding this to winemac.drv, since no known apps depend on this.
>
> Chip

Yes, wined3d currently depends on the implicit swapchain in a lot of
places and needs quite a bit of general cleanup for that first.
Activating the GL context without a drawable isn't going to work on
Windows and while wined3d on Windows isn't the most important target
we try at least not to break it. It shouldn't be necessary either, we
could e.g. create a dummy pbuffer in wined3d and make the GL context
current on it.

About the related winemac.drv patch, I guess it would be better to
revert it, yes. Sorry that I didn't realize about its implications
earlier. We could in principle change the test with some kind of
"todo_wine_if" shenanigans instead but that doesn't seem preferable.