Bug#446123: bugfix for endianness?
You're a genius. Gnome 2.20 had somehow switched out of the compositing cursor, which was probaly why I had suddenly noticed the change. When I changed to an argb compositing one - problem solved. I agere with what you are saying about not holding back the RandR functionality. Sometimes there is no elegant way forward in these situations. With best wishes, Adam On Jan 12, 2008 11:53 AM, Michel Dänzer <[EMAIL PROTECTED]> wrote: > > On Fri, 2008-01-11 at 20:10 +, Adam Bartley wrote: > > Ok, I can see why that is good coding practice on one level, but on > > the other hand a piece of code has been out in January 2008 which had > > been shown to be faulty months before in 2007. Maybe it should have > > been held out of circulation until the authors can come up with a > > solution? > > Ideally; the problem being that when RandR 1.2 support was added to > xserver 1.3, only xf86-video-intel had the corresponding driver support, > and that driver can physically only work on little endian platforms. The > issue was only discovered when RandR 1.2 support was added to the > xf86-video-ati radeon driver and working on PowerMacs, and now it would > be harsh to rip out the server support again for a minority. > > Note that only legacy two-colour cursors are affected, not modern 32 bit > ARGB cursors. > > Also, I think the old driver specific conversion code that was used > before RandR 1.2 could still work now; reinstating that would be a > relatively simple task for someone who is determined about fixing this > issue. > > > -- > Earthling Michel Dänzer | http://tungstengraphics.com > Libre software enthusiast | Debian, X and DRI developer >
Bug#446123: bugfix for endianness?
On Fri, 2008-01-11 at 20:10 +, Adam Bartley wrote: > Ok, I can see why that is good coding practice on one level, but on > the other hand a piece of code has been out in January 2008 which had > been shown to be faulty months before in 2007. Maybe it should have > been held out of circulation until the authors can come up with a > solution? Ideally; the problem being that when RandR 1.2 support was added to xserver 1.3, only xf86-video-intel had the corresponding driver support, and that driver can physically only work on little endian platforms. The issue was only discovered when RandR 1.2 support was added to the xf86-video-ati radeon driver and working on PowerMacs, and now it would be harsh to rip out the server support again for a minority. Note that only legacy two-colour cursors are affected, not modern 32 bit ARGB cursors. Also, I think the old driver specific conversion code that was used before RandR 1.2 could still work now; reinstating that would be a relatively simple task for someone who is determined about fixing this issue. -- Earthling Michel Dänzer | http://tungstengraphics.com Libre software enthusiast | Debian, X and DRI developer
Bug#446123: bugfix for endianness?
Ok, I can see why that is good coding practice on one level, but on the other hand a piece of code has been out in January 2008 which had been shown to be faulty months before in 2007. Maybe it should have been held out of circulation until the authors can come up with a solution? On Jan 11, 2008 6:21 PM, Michel Dänzer <[EMAIL PROTECTED]> wrote: > > On Fri, 2008-01-11 at 17:52 +, Adam Bartley wrote: > > I noticed on the freedesktop.org bugerport that a fix for this problem > > was *not* going to be released. is this true? > > Yes, because it's not a 'fix' but two hacks which together happen to > work. Nobody including the author(s) of the broken code seems to > understand how it's supposed to work in an endianness neutral manner. > > > -- > Earthling Michel Dänzer | http://tungstengraphics.com > Libre software enthusiast | Debian, X and DRI developer >
Bug#446123: bugfix for endianness?
On Fri, 2008-01-11 at 17:52 +, Adam Bartley wrote: > I noticed on the freedesktop.org bugerport that a fix for this problem > was *not* going to be released. is this true? Yes, because it's not a 'fix' but two hacks which together happen to work. Nobody including the author(s) of the broken code seems to understand how it's supposed to work in an endianness neutral manner. -- Earthling Michel Dänzer | http://tungstengraphics.com Libre software enthusiast | Debian, X and DRI developer
Bug#446123: bugfix for endianness?
I noticed on the freedesktop.org bugerport that a fix for this problem was *not* going to be released. is this true? I am not the world's most confident builder/compiler, and don't fancy trying to build these packages from source, after patching same, if help is around the corner. Regards, Adam Bartley
