"Derek Warren" <[EMAIL PROTECTED]> wrote:

> A `pstack core` from previous crashes showed this:
> core 'core' of 3745:    /usr/openwin/bin/Xsun :2 -nobanner -auth 
> /var/dt/A:2-G1aypb -nobanner
>   d18a0025 pix_encode () + 390

Ouch.  That's in the Sun Ray DDX shared object that gets loaded into
Xsun.  The instruction at pix_encode()+390 is a 'cmovl', which is a 
relatively recent addition (it came in with the Pentium Pro) to the 
x86 instruction set.  What kind of processor is on this mini-ITX 
board?  I bet it doesn't understand that instruction.

> I was afraid of that!  Thanks for pointing me in the right 
> direction. Here's the truss output (310k; 8.2MB decompressed): 
> http://derek.trideja.com/lib/xsun_crash-itx.txt.bz2

Sure enough, the last thing the Xsun process (2824) does is to blow 
up with an 'Illegal Instruction' fault at that address:

  5455  2824:   sigaction(SIGILL, 0x08044610, 0x00000000)       = 0
  5456  2824:   setcontext(0x08044480)
  5457  2824:       Incurred fault #1, FLTILL  %pc = 0xD18A0025
  5458  2824:         siginfo: SIGILL ILL_ILLOPC addr=0xD18A0025
  5459  2824:       Received signal #4, SIGILL [default]
  5460  2824:         siginfo: SIGILL ILL_ILLOPC addr=0xD18A0025

Unless there's a bugin the compiler (that's unlikely) we're probably 
compiling the Sun Ray DDX with options that allow the compiler to 
generate instructions that are too modern for this CPU.  I'll have
to go figure out whether this is intentional.  It feels like a bug
to me, if Solaris 10 is happy on this hardware then SRSS should be
happy to run there too.

OttoM.
__
ottomeister

Disclaimer: These are my opinions.  I do not speak for my employer.


-- 
___________________________________________________________
Sign-up for Ads Free at Mail.com
http://promo.mail.com/adsfreejump.htm

_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users

Reply via email to