priority 261251 normal
merge 261251 216933
thanks

On Sat, 2004-07-24 at 17:17 +0200, Emmanuel Fleury wrote:
> 
> I am hunting a very strange bug (probably hardware related). This bug
> has been reported by several people in bug #234556 as a bug of the
> xlibs package but I think I managed to push it down to the xserver
> package now (at least I believe so).

It's still the same bug though, why are you reporting it as a new one?
Merging with the existing bugs about this problem...

> This bug seems to appear only on architectures which contain a
> processor Transmeta Crusoe and/or a graphic card ATI Radeon Mobility
> (mine is a Radeon Mobility M6 LY). I should emphase the fact that we
> don't know which one from the Crusoe or the Radeon is responsible of
> this problem (in fact, I don't even know if it is hardware).

It's certainly not the graphics chip, as the code which generates the
error isn't even near the driver, let alone the hardware, and people are
seeing the problem with a variety of graphics chips, but only with
Transmeta CPUs.


> I compiled an Xserver with the debug options in order to follow what
> was going on inside the Xserver. I reached the specific mode where the
> bug was occurring and ran the Xserver with debug and... it worked. I
> mean the Xclient connected normally to the Xserver with debug. Of
> course, I checked that I was still into the specific mode by using a
> non-debug Xserver and the problem was still here.
> 
> At first, I really though I did something wrong while the compilation,
> so I took the Xserver package with debug options and it was the exact
> same behavior (i.e. once in the specific mode the XFree86 was crashing
> all the Xclients and the XFree86-debug was allowing the Xclients to
> connect).

Interesting. So it sounds like the optimized code either contains some
instruction(s) that work slightly differently on Transmeta CPUs than on
other x86 CPUs, and/or it triggers a bug in the code morphing.

Can others seeing the problem confirm that it doesn't happen with the
XFree86-debug server from xserver-xfree86-dbg?


-- 
Earthling Michel Dänzer      |     Debian (powerpc), X and DRI developer
Libre software enthusiast    |   http://svcs.affero.net/rm.php?r=daenzer



Reply via email to