Alan,

Thanks for the feedback.

Alan Coopersmith wrote:
> [cc'ed xwin-discuss, since most of these are X issues - for folks seeing it
>  there, the code review I'm responding to can be seen at:
> http://mail.opensolaris.org/pipermail/jds-review/2009-February/003857.html
>  mostly in the compiz-by-default script attachment at:
> http://mail.opensolaris.org/pipermail/jds-review/attachments/20090204/216bee31/attachment-0003.ksh
> ]
> 
>>     if [ "x$arch_type" = "xi86pc" ]; then
> 
> Systems running Xen/xVM aren't compiz compatible?

I don't have Xen/xVM systems to hand to test

> % /usr/bin/uname -i
> i86xpv

If they are compatible, then the check can be changed to :
    if [ echo $arch_type | grep -q "i86" ]; then

I need to verify this.

>>        verbose "Checking for Intel Graphics Card Type : "
>>         # List taken from SUNWdrmr/postinstall
>>         # Confirmed list (29/01/09)
>>         #   pci8086,2562 - compiz fails, driver problem
>>         #   pci8086,27a2 - compiz fails, driver problem
>>         #   pci8086,2a42 - compiz runs
>>         intel=`echo $prtconf_pv_output | $NAWK '{               \
> 
> I still think it's a mistake for your script to hardcode a list of
> Intel & ATI chip ids, since you'll have to keep up with new ones as
> they're added.

This is what the driver team recommended to Erwann, and thus this is what I
have implemented. I've contacted the DDU team to determine how the Device Driver
Utility picks up the video driver, and if this is easy and quick to implement I
may go down that route.

Agreed that the list would require maintaining for intel chips at least. Do 
packages install a copy of "postinstall" anywhere on one's system ? If so I
could just parse the file directly if installed.


>>     verbose "Check Xorg Log : $xorg_log,  for driver for graphics card 
>> $CARD_TYPE : "
>>         if $GREP -q "Loading $driver_path/nvidia_drv.so" $xorg_log; then
> 
> The contents of Xorg.0.log are not a stable interface, and can change without
> warning at any time.   "Loading" messages in Xorg.0.log do not indicate which
> driver is in use, just that the module was opened - it could have failed to
> attach to the hardware and unloaded.

Agreed the interface is unstable, do you think checking modinfo output is 
sufficient for this ? Ubuntu are relying on the Xorg.0.log check (Which dosen't 
mean it's right), but have been doing so since mid 2007.

Again feedback from the DDU team should also help here.


>>     if [ -f $COMPOSITE_INCLUDE ]; then
>>         major=`$GREP "COMPOSITE_MAJOR" $COMPOSITE_INCLUDE | $NAWK '{print 
>> $3;}'`
>>         minor=`$GREP "COMPOSITE_MINOR" $COMPOSITE_INCLUDE | $NAWK '{print 
>> $3;}'`
> 
> The version reported in the header does not tell you anything about what the
> server supports, and the header will not be installed on most OpenSolaris
> systems, since the header packages are not included on the live CD, nor are 
> they
> on SXCE systems installed with "End User" instead of Developer or Full 
> clusters.
 >
> You need to run something that calls XCompositeQueryVersion() to get the 
> version
> information.  I don't know of anything off hand included in the system already
> that does so and prints the result, but you could include a trivial helper
> program like the one attached.

Fair point, I thought the package SUNWxwinc was installed by default, if not
then I have no other choice but to call the API XCompositeQueryVersion() from a
helper binary (Thanks for that attachment), I was just trying to avoid extra
code/binaries being included

>>         verbose "   Horizonstal resolution ($hor_res) less than 
>> GL_MAX_TEXTURE_SIZE($max_texture_size)."
> 
> Typo in horizontal.

Well spotted :)

cheers

Matt

Reply via email to