Add log for verifyVoronoiTopology

[saeeedtorabi]

No description provided.

[pmconne]

This is an attempt to gather more info on this flaky failure, right?

[pmconne]

@pmconne This test generates 100 random points and had never failed until now. Since it's random, we cannot regenerate it. However, we added logs here #8639, so next time it failed we can regenerate it. Please note that I ran it 10 more times locally (i.e., for 10000 random points) and it never failed.

@saeeedtorabi the solution is to never use random data in tests - that makes them inherently flaky. You could of course generate the data randomly ahead of time and copy-paste it into your test, but the test should always execute against the same data.

[saeeedtorabi]

@saeeedtorabi the solution is to never use random data in tests - that makes them inherently flaky. You could of course generate the data randomly ahead of time and copy-paste it into your test, but the test should always execute against the same data.

ok @pmconne I will generate fixed points so the test is not flaky. However, we expect the test to pass for any random point and each failure shows a potential bug in the src code. Therefore, I keep random points but put them behind the flag enableLongTests so CI builds, which run with this flag disabled, don't fall over due to flakiness. However, we will need a dedicated CI job running geometry tests with this flag enabled so that the geom team can discover the rare edge cases like this. We would need someone knowledgeable to set this up for us.