Gerhard Niklasch on Sun, 17 Jan 1999 22:58:32 +0100 (MET)


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

Re: 2 bugs with pols in 2.0.12


In response to:
> Message-ID: <19990117161021.A14137@monk.mps.ohio-state.edu>
> Date: Sun, 17 Jan 1999 16:10:21 -0500
> From: Ilya Zakharevich <ilya@math.ohio-state.edu>
> 
> ? res=-23708160*x^9 - 225566208*x^8 - 1131314688*x^7 - 3968372736*x^6 - \
> 7276863744*x^5 - 1112932224*x^4 + 22694392512*x^3 + 49279399488*x^2 + \
> 50845741200*x + 22483386864;
> ? polgalois(res)
>   ***   segmentation fault: bug in GP (please report).
> 
> ? factornf(res,subst(res,x,y))
>   ***   bug in subresext, please report
> 
> (The second message is triggered even if we do not do polgalois().)

This is interesting... I began seeing the second phenomenon on my
old Linux box on one of the pre-2.0.12 snapshots if memory serves
(errm... it was present in 2.0.12.7),  but since Karim couldn't
reproduce it, and I couldn't reproduce it when building with -g [*],
we ended up blaming it on my ancient libc.  As far as I could tell
from running instrumented code, memory on the C stack was getting
corrupted even before subresext() was called.

Gerhard

[*] (maybe because of compiler hickups with inlining, maybe because
whatever causes the corruption interacts with the PARI stack allocation
machinery, which is implemented differently depending on whether we
build with -g or not)