2 fails with 'make check' in current CVS on IA32 Mac

[Thomas Treichl]

Appendum: Patch and ChangeLog entry.

   Thomas

Thomas Treichl schrieb:
> Hi,
> 
> I currently have two fails when doing 'make check' on the IA32 Mac platform - 
> but I think it is just a problem of accuracy?
> 
>  >>>>> processing /Users/Me/Development/octave/scripts/linear-algebra/condest.m
>    ***** test
>    N = 6;
>    A = hilb (N);
>    cA = condest (A);
>    cA_test = norm (inv (A), 1) * norm (A, 1);
>    assert (cA, cA_test, 2^-12);
> !!!!! test failed
> error: assert (cA,cA_test,2 ^ -12) expected
>   2.9070e+07
> but got
>   2.9070e+07
> maximum absolute error 0.00304392 exceeds tolerance 0.000244141
>    ***** test
>    N = 6;
>    A = hilb (N);
>    solve = @(x) A\x; solve_t = @(x) A'\x;
>    cA = condest (A, solve, solve_t);
>    cA_test = norm (inv (A), 1) * norm (A, 1);
>    assert (cA, cA_test, 2^-12);
> !!!!! test failed
> error: assert (cA,cA_test,2 ^ -12) expected
>   2.9070e+07
> but got
>   2.9070e+07
> maximum absolute error 0.00304392 exceeds tolerance 0.000244141
> 
>    Thomas
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: condest.diff
Url: https://www.cae.wisc.edu/pipermail/bug-octave/attachments/20071201/ea403f69/attachment.ksh 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: condest.changelog
Url: https://www.cae.wisc.edu/pipermail/bug-octave/attachments/20071201/ea403f69/attachment-0001.ksh