[Debichem-devel] Bug#1058454: Bug#1058454: prody: autopkgtest regression on i386

[Andrius Merkys]

Hi,

On 2023-12-12 13:48, Graham Inggs wrote:
> Source: prody
> Version: 2.3.1+dfsg-3
> Severity: serious
> Tags: sid trixie
> User: debian-ci at lists.debian.org
> Usertags: regression
> 
> Hi Maintainer
> 
> The autopkgtest of prody 2.3.1+dfsg-3 is failing when tested with
> numpy 1:1.24.2-2 [1].  I've copied what I hope is the relevant part of
> the log below.

[...]

> 345s E AssertionError:
> 345s E Arrays are not equal
> 345s E hessian is not symmetric
> 345s E Mismatched elements: 6 / 51984 (0.0115%)
> 345s E Max absolute difference: 4.4408921e-16
> 345s E Max relative difference: 7.83963072e-16
> 345s E x: array([[ 6.913751, 0.021868, 0.94014 , ..., 0. , 0. ,
> 345s E 0. ],
> 345s E [ 0.021868, 9.066915, -0.076779, ..., 0. , 0. ,...
> 345s E y: array([[ 6.913751, 0.021868, 0.94014 , ..., 0. , 0. ,
> 345s E 0. ],
> 345s E [ 0.021868, 9.066915, -0.076779, ..., 0. , 0. ,...

I think the issue here is with the tolerance of the equality criterion. 
Maximum absolute difference nowhere exceeds 1e-14, thus lowering the 
tolerance should with the failure. I am not sure how to adjust the 
frequency in this particular case, but I will give it a look when I have 
time.

Thanks,
Andrius