[Pkg-fglrx-devel] Bug#524871: Bug#524871: Issue list.

[Don Pellegrino]

On Tue, 2009-08-11 at 20:22 +0200, Patrick Matthäi wrote:
> The problem with fglrx is, that it is prop. and we can not "see" the 
> real reason for it. The invalid map handle error is there, but it should 
> just slow down the performance of opengl apps, maybe on some 
> cards/setups it also crashs.. I hope it will be fixed with the next release.

Thanks.  Is there any indication that AMD/ATI have accepted this as a
bug and that they are working on it?

> We are only supporting the configuration from the tree, so: we just 
> support etch fglrx with the etch xorg and etch kernel, same for sid and 
> testing, it is not possible to have multiple support for every stage here.

Right, I recognize the philosophy here although my experience with fglrx
from ATI has been that pinning back a few sid packages to testing or
unstable can make things workable.  This is relevant in the case where
fglrx can not be made compatible with the full set of sid packages.  I
presume it took some modification to get ATI fglrx release to work with
2.6.30 and I think it is worth considering when it is best to move fglrx
forward to the sid kernel or whether to give it a dependency on an older
(lenny / testing) kernel.  I don't know what guidelines might be
applicable for such decisions.

> Do not use the ati installer, if you realy feal annoyed by this issue, 
> downgrade your kernel and the fglrx debian version, but do not frickle 
> it around.

OK.  Thanks, that makes sense.  I ultimately found that fglrx-source is
unable to compile on the older kernel.  This is probably not too
important anyway as going with fglrx from ATI works on the older kernel
but also requires a downgrade of Xorg, which in turns breaks a lot of
GNOME components.  In the hopes that this may be a temporary issue until
the next fglrx release I went with the radeonhd driver for the 2400 HD
hardware.

Thanks for all the help and great job maintaining this package.