Philippe Elbaz-Vincent on Tue, 13 Sep 2005 00:19:45 +0200


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

Re: bug in qfminim() with pari 2.2.11 (current cvs) 

On Mon, 12 Sep 2005, Bill Allombert wrote:

The attached patch should fix that, but the doc is unclear whether 0
should be allowed or not, though I do not see any rationale for not
allowing it.




well it fixed the error, but the result is incorrect.
(patched version)
gp > qfminim(M,2*1512,0)
%2 = [22, 2940, [;]]
gp > B=qfminim(M,2*1512,22)[3];
gp > for(i=1,matsize(B)[2],print(B[,i]~*M*B[,i]);)
4313  <--- not possible
1680
1680
1680
1764
3053  <--- not possible
3053
3053
1512
1512
1512

a more likely result [Version 2.2.10 (development CHANGES-1.1169), x86 kernel]

gp > qfminim(M,2*1512,0)
%37 = [42, 2940, [;]]
gp > B=qfminim(M,2*1512,42)[3];
gp > for(i=1,matsize(B)[2],print(B[,i]~*M*B[,i]);)
2268
2268
2268
2940
2940
2940
1680
1680
1680
1764
1764
1680
1680
1680
1764
1680
1680
1680
1512
1512
1512


Cheers, Philippe.