Buggy repeat modes in Render

Michel Dänzer michel at tungstengraphics.com
Thu Feb 28 03:51:30 PST 2008


On Wed, 2008-02-27 at 07:39 -0800, Carl Worth wrote:
> On Wed, 27 Feb 2008 12:54:36 +0100, Michel Dänzer wrote:
> > I'm afraid it's not feasible to derive this from a single number - any
> > breakage could be due to either of at least pixman, fb, the X
> > acceleration architecture or the driver.
> 
> Right.
> 
> But fb and the acceleration architecture are part-and-parcel with the
> X server code, so the one reported version number is fine there. 

You can't tell XAA from EXA (or GLucose, ...) though. AFAIR one of those
workarounds already penalized EXA for an XAA bug, it would be nice if
that kind of thing didn't happen again.

> And pixman can be similarly covered to a large extent, (assuming that
> people ship pixman and server bits together as "released").

Not sure that's a valid assumption, but maybe I'm just spoilt from using
Debian. :)

> The driver part is much harder to deal with since the X server doesn't
> advertise a name or a version number for the driver. Could we perhaps
> just fix that?

I'd rather not - as someone else mentioned, I think the idea of clients
behaving differently depending on driver version is rather horrid.

> Meanwhile, the reports I've gotten on this particular bug suggest that
> there is at least some element to it that is not driver specific.

Still, I suspect there's some way to go until the common drivers handle
all of these cases correctly.


> > Would it be feasible for Cairo to run tests like these on startup and
> > not use the features that it finds to be broken?
> 
> I don't think so.
> 
> We've already got a handful of workarounds in cairo for X server bugs,
> and we really only expect that number to go up, not down. So doing
> anything more complex than checking version numbers will quickly start
> taking a non-trivial amount of time.
> 
> Particularly since doing work at "cairo startup" would mean slowing
> down the startup of every graphical application. Not an appealing
> prospect.

Maybe the results could be cached, e.g. by storing them in atoms.

This is the same philosophy used by the autotools - in order to find out
if something works, just try it instead of guessing from other things.


-- 
Earthling Michel Dänzer           |          http://tungstengraphics.com
Libre software enthusiast         |          Debian, X and DRI developer




More information about the xorg mailing list