Re: [Gcl-devel] RE: [Maxima] modulus

[Camm Maguire]

Greetings, and thanks for the report!

"Stavros Macrakis" <address@hidden> writes:

> > > in the GCL version, there is a bug where huge 
> > > integer calculations corrupt memory.
> 
> > Am unaware of this.  Please file as detailed a report as 
> > possible if you believe this to be a GCL bug.
> 
> In a fresh GCL 2.5.0, (expt 234 234234) immediately crashes with an
> illegal instruction (0xc000001d) at location 0x0053f611.
> 

OK

1) Please try with latest gcl for windows, at

http://www.cs.utexas.edu/users/boyer/gcl
or
http://people.debian.org/~camm/gcl

2) I suspect that it might not help, that the heap is growing into
   some other used memory area (on Windows only), and that this is
   also the cause of the failure to build pcl from source (on Windows
   only).  Mike, could you please recap where DBEGIN is, where the C
   stack is and in which direction it increments, where the static
   areas are, and where any shared libs used by our gcl image are
   mapped.  This would be most helpful.

3) On my i386 Linux gcl, this does not quit, does a few finite gbc
   calls to get enough room (i.e. no memory overflow), but then hangs
   for a very long time trying to calculate the answer.  I cannot tell
   if this is a problem, or whether the number is effectively too
   large to be processed in a reasonable amount of time.

Take care,

> In a fresh Maxima 5.9.0 built on GCL 2.5.0, 234^234234 will return
> (though I have not checked whether the result is correct), but if you
> repeat the command a couple of times, you get an error:
> 
> (C1) 234^234234$
> (C2) 234^234234$
> (C3) 234^234234$
> Error: Caught fatal error [memory may be damaged]
> Error signalled by EXPTRL.
> 
> After this, other things go wrong,
> 
> rat((x+1)^20);
> Error: Caught fatal error [memory may be damaged]
> Error signalled by PEXPT.
> 
> Sometimes I think it crashes the process completely, but I haven't been
> able to reproduce that right now.
> 
> Though 234^234234 is large, it is not *that* large -- about 60,000
> 32-bit words.  And if the problem is some sort of memory allocation
> overflow, it should give a relevant error, not just randomly fail.
> 
> Perhaps this has been fixed in a more recent GCL...?
> 
>        -s
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah