It fixes my test case, that is starting up with 32 bits depth console and
then switching to 16 bits depth -
since the depth is cached, xf64-video-vmware gets 565-weight but 32 bit
depth, and refuses to start.
There is another similar bug I'm tackling now - when the window resizes,
qemu_console_resize() creates
a new console with a hardcoded 32-bit depth, instead of current depth,
which breaks the display.
I'm gonna test the live migration too.
Alex
On Wed, Apr 3, 2013 at 12:49 PM, Gerd Hoffmann <kra...@redhat.com> wrote:
> On 03/29/13 17:28, Alex DAMIAN wrote:
> > From: Alexandru DAMIAN <alexandru.dam...@intel.com>
> >
> > Do not cache depth and bypp information in the device state.
> >
> > This resolves a bug where Xorg video-vmare driver refuses
> > to start up because the depth value read is the one cached from the
> > device start (default 32 from ui/console.c) and it is not consistent
> > with the graphical console depth, which may be different from
> > the default depth.
>
> Does it actually work? /me posted a simliar patch and according to Jan
> (Cc'ed) it doesn't fix the issue.
>
> > @@ -1113,7 +1111,6 @@ static const VMStateDescription
> vmstate_vmware_vga_internal = {
> > .minimum_version_id_old = 0,
> > .post_load = vmsvga_post_load,
> > .fields = (VMStateField[]) {
> > - VMSTATE_INT32_EQUAL(depth, struct vmsvga_state_s),
>
> This breaks live migration.
>
> cheers,
> Gerd
>
>
>
