Re: Status of SMP on sbus machines?

[Chris Ross <cross+netbsd%distal.com@localhost>]

On Oct 19, 2008, at 20:18, Chris Ross wrote:
> On Oct 8, 2008, at 13:10, Martin Husemann wrote:
> > On Wed, Oct 08, 2008 at 01:06:21AM -0400, Chris Ross wrote:
> > > this resolved?  If I built a recent current and tried it on a dual-
> > > Ultrasparc SBUS machines (Ultra-2), would it potentially work?
> >
> > Yes, it was solved. I'm using a dual U2 quite heavily.
>
> Excellent.  Thank you.
>
> I'm trying to run my dual-UltraSPARC (UE2, with UltraSPARC-I's in  
> it), and while it seems to generally run alright in multiuser, I'm  
> getting the following early in the process of trying to build pkgsrc/ 
> lang/perl5 (5.10.0) on it:
>
> trap: textfault at 200fa0!! sending SIGILL due to trap 16: illegal  
> instruction
>
> This happens when running the "try" binary, which compiles okay, but  
> fails when run.  I've tried compiling what it claims to be the  
> content, with what it claims to be the compiler line, and I don't  
> see this.  So, something's "odd".  But, I'm wondering if that error  
> message might mean anything to someone else...

  Still more information.  I did eventually get perl to build.   
Actually, after the configure phase, it got much better.  And, I once  
got it to complete configure without more than one or two crashes like  
the above.

  So, this morning, I tried a "make test" on the built perl.  While  
perl is running, *it* sometimes will throw this sort of error.  I got:

Oct 20 09:37:32 usparc /netbsd: trap: textfault at 41502fc0!! sending  
SIGILL due to trap 16: illegal instruction
Oct 20 09:39:42 usparc /netbsd: trap: textfault at 41a02fc0!! sending  
SIGILL due to trap 16: illegal instruction
Oct 20 09:39:44 usparc /netbsd: trap: textfault at 41a02fc0!! sending  
SIGILL due to trap 16: illegal instruction
Oct 20 09:39:45 usparc /netbsd: trap: textfault at 41502fc0!! sending  
SIGILL due to trap 16: illegal instruction

  These were generated by the tests:

t/io/crlf
t/op/inccode-tie
t/op/inccode
t/op/incfilter

  ...and I stopped the tests shortly after that.  Unexpectedly (to  
me), these seem to be consistent. Running "make test" again produced  
the same failures.

  If anyone has any notions, or any suggestions as to what I could  
try next to help track down the problem, let me know.

  Thanks...

                             - Chris