Test failing due to the trivial differences of some double values (e.g. Apple processor vs. Intel processor)

[luchengqi7]

In some occasions, the double values generated on Apple Silicon processor are slightly different from those generated on a Linux/Window machine with Intel processor. This may lead to problems in some integration tests, where output plans are compared to the expected output plans. The scores of the plan may be trivially different (i.e. below 10e-6), and the test will fail (a recent example: PR #2997).

This problem can be sometimes frustrating, especially for new users, as local tests may fails while the remote test passes (or the other way around).

As suggested by @michalmac, one solution is to update the way how we check the equality of two plans file. Are there any suggestions from the community?

[kainagel]

I had the same problem. Unfortunately, if you look into PopulationUtils#comparePopulations(..) (which points to PopulationUtils.equalPopulation(...)), I have no idea how to do this in a simple way. I don't even completely understand how this works.

[steffenaxer]

Same for me!