Re: pari-2.0.10.beta

[Gerhard Niklasch <nikl@mathematik.tu-muenchen.de>]

In response to:
> Message-ID: <Pine.GSO.3.96.980712234056.21512A-100000@abello>
> Date: Sun, 12 Jul 1998 23:54:03 -0400 (CST)
> From: "Matias Atria (F. de Ciencias Depto Matematicas)"
> 	<matria@abello.dic.uchile.cl>
> 
> Using pari-2.0.10 I get the following:
> 
> ? ?inth0
> inth0() = r1/n-inth(h)
[...]
> ? 1.0 - inth0()
> %28 = -0.973140116944
> ? 1 - inth0()
>   ***   overflow in I+R
> ?
> 
> I'm running the SuperSparc 32-bit version on a Sun Sparc20. If you need more
> information just let me know.

Weird;  on a Pentium under Linux, I get the overflow message when the
first summand is a real  (and I don't get it when the first summand is
an integer).  Moreover, the special shape of your function seems to have
nothing to do with it;  it seems to be triggered by any argument-less
user-defined function in this syntactic position.

Apparently something is kicking the parser off track here...

And badly so:
---8<---
(06:58) gp > mysqrt()=sqrtint(1);
(06:58) gp > ?mysqrt
mysqrt() = sqrtint(1);
(06:58) gp > 1.-mysqrt()
  ***   the PARI stack overflows !!!

  ***   Warning: doubling stack size; new stack = 8000000.
(06:58) gp > 1-mysqrt()
prec = [5, 16]
%25 = 0
--->8---
(If you've never seen that `prec' line before, it's a side effect
of cranking up the default(debuglevel,_) to a high value.  Unfortunately,
the parser doesn't produce diagnostic output.)

I'll look into this later today;  Karim's mail server at Orsay seems to
be down  (or at least it was down for several hours)  so he may not have
got this yet.  Thanks for the report, it's indeed a new phenomenon
(to me at least!).

Gerhard