Test PMQRTest>>testDecompositionOfMatrixCausingErraticFailure is failing since Pharo 13

[SergeStinckwich]

Apparently something change since Pharo 13, as the test PMQRTest>>testDecompositionOfMatrixCausingErraticFailure move from green to red.

@hemalvarambhia as you have been working previously on this code, maybe you have an idea of why this test is failing now in Pharo 13 and Pharo 14: https://github.com/PolyMathOrg/PolyMath/pull/288/commits

I'm not sure to understand the intention of the test:
self assert: pivot equals: #( 3 4 3 nil )

Now what is returned since Pharo 13 is #(3 4 3 4).

[SergeStinckwich]

We are planning to work with @hemalvarambhia on PMQRTest>>testDecompositionOfMatrixCausingErraticFailure.
The test is failing for the moment. I put an expectedFailure in order to have all our tests green on the CI.
We need spend sometimes to understand how QR Decomposition is working.

[DurieuxPol]

Hello ! I checked the QR decomposition, the two methods to decompose, and the failing test.
First, I'm not sure why we have two methods to decompose, one "simple" and one with pivot. I don't see the point of the latter and if they are even really different from one another.
And also about the failing test, it's kind of weird because it decomposes a matrix, then expects that q*r gives a matrix different from the initial one. And it expects that there is a nil in the pivot for some reason.
I think I will try to refactor and improve the implementation here.

[DurieuxPol]

The issue is that there is a call to closeTo: in PMQRDecomposition>>#decomposeWithPivot, which relies on the Float DefaultComparisonPrecision that changed between Pharo 12 and Pharo 13 (0.0001 to 1.4901161193847656e-8).
To fix this I will call closeTo:precision: instead, with a customised precision for this class.