mpz_mul core dump - attempt 2

[Kate Minola]

Alex wrote:

> I built and ran the code and it worked fine.  I used
> gcc3-3.4.1 to build your code.
> 
> However, my libgmp and gcc3 is from blastwave.org.  I know
> that gmp is built with Sun's compiler and I am fairly
> certain that gcc3 is also.  So, I suspect a bug in
> gcc3-3.4.1.
> 
> I am using the stock stack size of 8192kb.  See the
> following.  Are we using the same hardware?

My understanding is that an Ultra-60 is an "ultrasparc2"
as returned by gmp-4.1.3/config.guess.

I have built gmp-4.1.3 with both cc and with the recently
released gcc-3.4.2 on an ultrasparc2.  I still have the
problem.

Torbjorn wrote:
  
> I don't have any system where I have been able to
> reproduce this problem.

Can you tell me the identity of one sparc system (identity
as returned by gmp-4.1.3/config.guess) that you have
tried?  And what compiler you used to build gmp-4.1.3 on
that system?

I have been going around and seeing what sparc systems
I have access to.  So far I have found
 
ultrasparc-sun-solaris2.7
ultrasparc2-sun-solaris2.8
ultrasparc2i-sun-solaris2.7
ultrasparc2i-sun-solaris2.8
ultrasparc3-sun-solaris2.8
ultrasparc3-sun-solaris2.9

Do we have any machines in common?

> Somebody else need to debug this.  How about that you
> debug it yourself?

No argument.  But before I dig too much further into
the code, it would be helpful to know whether I am hunting
a gmp bug ... a gcc bug ... an OS bug.  That is why I 
would like to either get the bug confirmed or find a sparc
system on which the problem does NOT occur.

Kate Minola
University of Maryland, College Park