atom boards with i915 video

gene heskett gheskett at
Mon Jan 16 08:15:57 PST 2012

On Monday, January 16, 2012 10:19:36 AM James Cloos did opine:

> >>>>> "gh" == gene heskett <gheskett at> writes:
> gh> ... but from the results obtained when I try to use it, the i915
> gh> driver hasn't those capabilities.
> You may want to try with:
>     :; export LIBGL_ALWAYS_SOFTWARE=1
> that avoids the hardware opengl; mesa's software renderer may have what
> you need, even if the i915 hardware driver does not.  Given your usage
> description, it would seem that the possibly lower rendering speed would
> not be an issue.

Not a bit, as long as it can render it in a few seconds time, visually 
reasonably accurately.  It isn't that visual output that counts, its the 
100k or more worth of GCode it spits out when 'make path' is clicked on and 
then a click on the save ngc button.  That will be the file fed to the 
milling machine, and which may takes several hours to run per board side 
for a complex pcb.  These simple boards won't even use the 2nd side copper 
other than to chamfer the drill holes so a passing component lead won't 
inadvertantly short to it.  It is relatively important the board won't warp 
so dbl sided board is being used.  If I can get the measurements correct, 
the mounting brackets might even be soldered to the top side copper.

>     :; env LIBGL_ALWAYS_SOFTWARE=1 glxinfo

SSH -Y'd into that box from here, I suspect the answers are coming from 
this box as the display is named as 11:0. Worthless IOW. :)

> will show what your version of mesa supports in software mode.

Quite extensive, but likely bogus from this console, 125 feet of cat5 & 2 
hubs away.

> -JimC

We finally did make it work on this box, seems the real problem is java and 
its famous 'write once; pray everywhere' reputation.  I straced it, and put 
links to what it was looking for at the first location it looked for the 
libraries it needed, which sped up the failure message by several seconds, 
but didn't effect the message.  According to the strace, it was looking 
without success at several quite lengthy paths, then looking at ldconfig's 
output file and repeating that same fruitless search so that each needed 
library's search was about 45 lines of output on the console.  Screw it, 
put a link in the first place it looked, although I did have to do a mkdir 
-p with that path about 7 layers deep just to get to that 'first' place it 
looked.  Not a very good way to run a railroad IMO.

I presume strace was not following all the library needing another library 
calls.  So one of the guys on the emc list refactored the .jar to actually 
contain all that stuff, and we wrote a script we put in /usr/bin to export 
the LD_LIBRARY_PATH it needed, and now it runs just fine on this box with 
its lower range nvidia card and their driver.  I've since found a couple of 
minor buglets that are not show stoppers in visolate.jar, reported on the 
mailing list at MIT, but over this 3 day weekend, no replies have been rx'd 
yet.  Maybe tomorrow.

I suspect a similar fix, with the new .jar, and your example line above 
will make it run on that Atom/Ubuntu-10.04 LTS box too.  But with temps in 
the teens, I have not gone out to the shop to try it from that machines own 
consoles.  Diabetic's shouldn't temp fate (and frostbite) as it might cost 
them their feet.

That script, for that machine would then look something like this:

# change following path to the location of the java3d *.so files
export LD_LIBRARY_PATH=/etc/alternatives/jre
# change following path to the location of visolate.jar
java -jar /usr/share/java/visolate.jar

Something like that anyway. :(  The second export line isn't needed here of 

Now if scp's syntax wasn't so darned opaque, I might be able to get it 
copied over and similarly setup on that box.  I might have to get better at 
the praying part though.  It seems to be a requirement for successful java 
usage. :(

Thank you very much, James.

Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
My web page: <>
<Addi> seems to have replaced one of its routers with a zucchini.

More information about the xorg mailing list