Karim Belabas on Wed, 26 Nov 1997 19:51:13 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: pari-2.0.alpha/linux-i586


> Clemens Heuberger <cheub@weyl.math.tu-graz.ac.at> wrote:
>
> I have just compiled pari-2.0.alpha on a Pentium 200 running Linux
> (2.0.32). Do I have to worry about the following?

Yes you do. I can reproduce it though (on a similar machine). I'll try to
understand what exactly is going on and post a patch (don't expect it tonight
though).

The polred & co things are bad enough already since it means something went
wrong (or at least unexpectedly) with the random number generators.

>   ? \q
> !   ***   segmentation fault: bug in GP (please report).

  And this is very bad since it means the stack was somehow corrupted using
the compatibility mode (which it shouldn't do since the functions ultimately
called and do the real possibly-stack-threatening work are the same). Am I
right in assuming that everything went smoothly in non-compatible mode ?
(that's what happened to me).


Some comments on Gerhard's comments on Clemens original mail:

> Gerhard Niklasch <nikl@pchelwig1.mathematik.tu-muenchen.de> wrote:
>
> > > PROBLEMS ARISED. See out.sta.diff
> 
> Yes.  It should be `PROBLEMS HAVE ARISEN' or `PROBLEMS WERE NOTED'.

should it ? You'd like it better this way ?

> No.  polred and friends may produce stuff in different order, with
> different signs, and sometimes even different stuff. 

Not in the bench suite. This should be completely deterministic.

> (Karim, does the test suite do a setrand before this?  Perhaps it should.)

They do. Not formally, but GP is started in test mode (which is not
documented and should not be of any use to anyone not rewriting the
interpreter, but does in particular ensure that no user-configurable stuff
gets in the way. In particular the seed pari_randseed is set to 1 when the
bench begins).

   Karim.
--
Karim Belabas                          e-mail:
Max-Planck-Institut fuer Mathematik       karim@mpim-bonn.mpg.de
Gottfried-Claren-Str. 26               tel:
53225 Bonn (Germany)                      (00 49 228) 402-245