> Due to the recent work by Martin Bochnig to migrate XSun legacy SPARC
> video drivers to Xorg using snv_95, we want to review the migration of Xorg 
> 7.4
> into Solaris 10 and OpenSolaris 2008.11. This work is to FULLY replace
> XSun, if possible, and merge XSun into a truely compliant Xorg 7.4
> implementation with Sun's XSun "enhancements" for backward compatibility 
> purposes. 
> 
> This work involves video drivers like OpenChrome, Intel, ATI, and Nvidia
> driver compatibility with Xserver 1.5. Also, the input device drives like
> touchpads, keyboards, and tablets.
> 
> More later, as the Sun X team can provide the extensive details and
> history on how the project can achieve its mission.
> 
> ~ Ken Mays


Ken, thanks for requesting this. But the current stability may not be 
sufficient. I have a long list of things that still need to be addressed. This 
can only be done by somebody who has all the test hardware (ask me for details 
Sun, if you don't believe this).
Plus it is a fulltime job. But ....
And that's the problem.
The X11 group does not have the testing hardware, but doesn't recognize the 
value that my collection of boxes constitues.

This text is being written on a Blade 2000 with gdm enabled, cde-login 
disabled, and /usr/openwin pointing to /usr/X11. There are NO legacy Xsun 
/usr/openwin/lib libs on this machine anymore. It only has access to the open 
free fox-gate's fonts (the fonts look a bit worse, than normally on SXCE, but 
this can probably be tweaked at least a bit, like on Indiana). I'm running JDS 
here and writing this via webmail inside firefox 3. Note that I have renamed 
int10, because it crashes when you try to start glint on the Psycho of the 
Ultra30 with latest OBP. For this reason, and now comes something very 
important, if you want to use the Xorg server (with disabled int10) on a pci 
chipset, you will need to hand-edit xorg.conf and add a matching BusID section 
(you get the info from scanpci). Or rename back 
/usr/X11/lib/modules/libint10.so__ORIG to /usr/X11/lib/modules/libint10.so, 
because it does not crash on most other configurations. The U30 glint crash 
needs some mdb analysis when I have more time for such a not so common config.

p.s. Yes, this stuff behaves differently on every platform. Earlier today I had 
removed yesterdays gdm-compatibility patch from the fox-gate and have replaced 
it with a much smaller version. I had written that "I want to keep things 
simple, because that smaller version was all Xorg needed on sunffb". WRONG 
again: On Ultra30 "yes". On SunBlade 2000 "No". I need to re-create the 
archive, upload a new one and re-enter the old randr/rrxinerama.c patch into 
the fox-gate. Because right now you are not able to start gdm on the 
Blade1000/2000 (and maybe more platforms that I didn't test yet). gdm would 
always quit and would print "xinerama enables, but <= 0 screens?" to syslog. It 
is complex. You always need to test every change on all platforms, frame 
buffers, kernel /dev/fb console drivers, Xorg userland ddx modules. 

So there are still bugs and problems here and there that I only worked around 
for now. To debug and fix every of them costs time. But Sun doesn't want to 
sponsor this. Note that I would be happy with less than 1/10th of what this 
would normally have costed. But I cannot influence Sun's decisions. They don't 
want to pay a single dollar for all their loyal SPARC-Desktop users.

This is not a rant. It seems to be the situation.
I have added indiana-discuss to the CC because it is very relevant to 
SPARC-Indiana's outlook.


%martin

pp.s. As  mentioned, I'm loading up a new archive to the same location in 1 
hour, which will fix the Blade1000/2000 gdm problem. I will also rename back 
int10, so that unaware users have less problems (as it works with the majority 
of (semi-)supported pci cards vs. workstation_models.

Reply via email to