Testbot D3D failures (Re: [PATCH 6/6] ddraw: Avoid passing invalid device type to wined3d.)

[Henri Verbeet]

We already talked about this on IRC, but for the benefit of the mailing list:

On Mon, 8 Apr 2019 at 07:05, Zebediah Figura <z.figura12 at gmail.com> wrote:
> It seems that qxl is not the only driver affected by this: bug 33290 [3]
> describes a similar problem with NVIDIA binary drivers (those, at least,
> that actually report multiple modes, and so don't fall back to XRandR
> 1.0 in Wine regardless). This was diagnosed there too, in comment 15,
> and a patch was submitted, which received the comment [4] "Updating the
> screen size isn't necessarily wrong, although avoiding panning is the
> wrong reason." I can only imagine, then, that "because the spec says so"
> is the right reason, but I may be misunderstanding or missing any number
> of things.
>
The typical starting configuration is DFPs running at their native
resolutions, and applications, typically games, switching to a lower
resolution. In that scenario the screen size is always large enough to
contain the CRTCs. In fact, changing the screen size may cause desktop
environments to rearrange panels and icons. Depending on the specifics
they may of course already do that when changing CRTC sizes, and
rearranging things may or may not be considered desirable by the user.

We would of course need to set the screen size when changing to a
configuration that would be larger than the current screen size.

Another potential reason for us to update the screen size is making
sure SM_CXVIRTUALSCREEN and things derived from that match what they
would be on Windows. That's something that would need some tests, of
course.

As for panning, that's something that can be configured in xorg.conf,
or with e.g. xrandr. The initial state for panning may be different
between drivers.

Henri