Re: [Qemu-devel] Suspicious fall through in qemu_default_pixelformat()

[Stefan Hajnoczi]

On Tue, Nov 22, 2011 at 10:06 AM, Markus Armbruster <address@hidden> wrote:
> Case 24 falls through to case 32, which overwrites all of case 24's
> work.  Looks wrong.  Git blames commit 0da2ea1b "fix endianness problem
> sharing the videoram buffer".
>
> Patching in a break is easy enough, but since I don't actually
> understand how the function is used, I have no idea what exactly this
> fixes.  Stefano, can you help?

(I don't have an 'o' at the end of my name, but...)

The 24 bpp and 32 bpp formats are the same except 32 bpp adds an 8-bit
alpha component.  We can safely insert a break.

I'm confused about the alpha component though.  It seems we discard
alpha for 32 bpp.  It's unclear whether this is intentional or has
just never been noticed because we don't use alpha anyway.

Stefan