On Thu, 9 May 2002, Michael wrote:

>> >> High resolution video requires a lot of video memory.  DRI 
>> >> requires approximately four TIMES that much video memory to work 
>> >> properly.  However, people are using insane high resolutions like 
>> >> 1600x1200 and using DRI, then running Wolfenstein 3D, and it 
>> >> crashes.  Hmm, gee... I wonder why.
>
>Perhaps because the voodoo3 will happily run at these resolutions?
>Doesn't sound insane to me.

Might not sound insane, but the bare fact is that it does not
work in XFree86.  I could care less personally if it ever does,
as it is obsolete hardware.  I only care that XFree86 does not
crash.  If nobody else is interested in fixing it either, then I
believe the X/DRI sources should detect this problem and disable
it by default as I have done, unless we now consider it ok for
the X server to crash for no reason other than it being
configured in a way that the drivers do not support currently.

I am interested in hearing the opinions of DRI and XFree86 
project members on the this particular issue with the tdfx 
driver.  I have a feeling however that the DRI developers favor 
stability, reliability, and security over driver features, 
supported resolutions and the like.

If someone wants to modify a driver and use it in a known 
unstable manner, and they're lucky enough to not trigger problem 
behaviour, or to trigger it infrequently enough that it doesn't 
bother them, that is their perogative.  But it is not acceptable 
IMHO to knowingly ship a driver which crashes, and for which the 
reason for the crash can be easily fixed or worked around by 
detecting problem configurations and disabling features known to 
be unstable.

If someone wants to see this problem fixed in a better way than
just being worked around by disabling the known problematic
configuration, then it is a good project for someone to work on
as a personal itch to scratch, and to ensure that the tdfx
drivers continue to function and to access as many features of
the hardware as possible.  Unmaintained drivers eventually become
non-working drivers, and eventually become completely disabled
drivers.  As hardware becomes older, and nobody is funding
development of the code, there is less and less incentive to fix
bugs in the code, when there are fixed amount of human resources
available, and newer drivers for more modern hardware have their
own fare share of bugs needing fixing.  It makes the most sense
to spend resources first on the modern hardware, and much lower
priority on the older hardware.  It is in these situations when 
the open source development community needs to step in and 
scratch their personal itch by hacking on code and contributing 
it.  Not unlike the Linux kernel's own developmental processes.

Fortunately, due to the good work that the DRI guys, and others
have done in the past on the tdfx drivers, they are in fairly 
good shape now.  This one problem is the only major problem that 
pops up now that I can think of on Voodoo 3/Banshee hardware.

If someone is interested in solving this for real, let me know, 
and I will put the Voodoo 3 specs up and provide a URL to them.  
Also, it would be a good idea to join the #dri-devel meetings on 
irc.openprojects.net on Mondays at 4pm EST to discuss how one 
might go about finding/fixing the problem.  I'm sure that Daryll 
Strauss, and others who have worked on this code before, will be 
glad to help out anyone interested by providing pointers.

But until such bugs are truely fixed, stability trumps whiz-bang.



-- 
Mike A. Harris                  Shipping/mailing address:
OS Systems Engineer             190 Pittsburgh Ave., Sault Ste. Marie,
XFree86 maintainer              Ontario, Canada, P6C 5B3
Red Hat Inc.
http://www.redhat.com           ftp://people.redhat.com/mharris


_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert

Reply via email to