Opera and EXA: ``X Shared memory extension is not available''

[Keith Packard]

On Mon, 2008-06-23 at 02:17 +0200, Juliusz Chroboczek wrote:
> > It looks like EXA disables shared pixmaps on purpose:
> 
> It looks like you agree with Mark Vojkovich on this point, then.
> I don't claim I understand the issue, but I'm more than willing to
> trust the two of you.

Shared memory pixmaps were interesting when all pixmaps lived in main
memory - it was trivial to share them with applications, and you could
do some fun non-X drawing as a result. With pixmaps wanting to live in
the graphics engine, it isn't practical to share them with applications
as that would require all kinds of synchronization between the
application and the GPU. So, now MIT-SHM pixmaps end up in main memory,
drawn with the CPU.

Also, sometimes depth-24 frame buffers are stored as 24bpp, but so many
applications had bugs with their 24bpp image handling code that we gave
up and exposed these screens as 32bpp and did magic image format
conversion for GetImage and PutImage. Exposing shared memory pixmaps for
these screens means that we now have pixmaps at 32bpp when the hardware
is running at 24bpp. Icky.

I'm still not sure why MIT-SHM pixmaps was disabled in EXA; it should
all work as we've dealt with these problems for years, but it is true
that not providing them offers applications a more realistic view of how
the system works -- drawing to MIT-SHM pixmaps in the X server is no
faster than drawing to them in the application, and probably a lot
slower.

> > Opera shouldn't be complaining about their absence.
> 
> So are you positive nothing else is weird about MIT-SHM and EXA?

Not that I know of. And this combination is used rather a lot these
days...

> I need this confirmation before I file a bug against Opera.  (You
> don't fool with Norwegians just like that, they're a tough bunch.
> They eat Moose for breakfast.)

Opera certainly shouldn't complain about the lack of MIT-SHM pixmap
support; it's always been optional. They should deal with it by using
MIT-SHM images instead.

If Opera requires this to work, it shouldn't be that hard to turn on
again.

-- 
keith.packard at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.x.org/archives/xorg/attachments/20080622/50d6a400/attachment.pgp>