Re: [Qemu-devel] [RFC] qemu VGA endian swap low level drawing changes

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Sun, 2014-07-06 at 12:19 +1000, Benjamin Herrenschmidt wrote:

> I've started to look (and while at it added use of the dirty bitmap to
> catch changes to the HW cursor image just because it looked easy).
> 
> One obvious issue: Your patch:
> 
> gtk: update mouse position in mouse_set()
> 
> Completely breaks cursor movement in NT4 (I haven't checked with other
> guests). It works fine without the patch. This is after cherry-picking
> on top of my series on github.
> 
> Now the cursor is still a white rectangle :-) I need to investigate that
> one a bit more.

Ok, I've found the cirrus bug that causes that white rectangle, but also
the broken icons in 8bpp mode.

We're basically tripping that test in cirrus_bitblt_rop_fwd_*

    if (dstpitch < 0 || srcpitch < 0) {
        /* is 0 valid? srcpitch == 0 could be useful */
        return;
    }

Because when called from cirrus_bitblt_cputovideo_next() we
always pass 0 for both pitch (we do one line at a time).

That done, we now get both HW "emulated" and HW "qemu" cursors
working in 8bpp.

The cursor is still absent in 16, probably a minor glitch, I'll
try to figure that out next.

Cheers,
Ben.