fix a bug in AR picker

[megies]

What does this PR do?

Presumably fixes buggy AR picker C code. Fix is done by @jwassermann.

I'll open it against master for inclusion into 1.2.0 since 1.1.1 is already in release candidate stage and also because this might change results significantly (just guessing, really).

It's a bit scary that the AR picker test doesn't even need to be adapted, but hey that's life. Test was adapted/improved.

PR Checklist

• Correct base branch selected? master for new features, maintenance_... for bug fixes
• All tests still pass.
• Any new features or fixed regressions are be covered via new tests.
• Any new or changed features have are fully documented.
• Significant changes have been added to CHANGELOG.txt .

CC @jwassermann

[krischer]

Any chance to add a regression test? @jwassermann do you have some data that originally triggered the bug?

[megies]

Any chance to add a regression test? @jwassermann do you have some data that originally triggered the bug?

Sure we can, the change is pretty subtle though, the result is changed for the S pick only and also only slightly. In the test case from 32.2s to 32.1s or so. I'll make the test more fine-grained again, there's a strange comment in there that the result is very machine dependent and therefore it's rounding to whole seconds in the test..

[krischer]

👍 Under the condition that CI and also the docker tests pass. The circle CI failures are dealt with in #2158 .