Intermittent segfault in pypy

[jpivarski]

Version of Awkward Array

HEAD

Description and code to reproduce

We now have a test that's tracking it, but that test fails (with the segfault) occasionally, and we eventually need to do something about it.

If we knew that the segfault is only in pypy, then I'm partially inclined to just declare that we don't support pypy. (How important is it? Who uses it?) But, it's not uncommon for a segfault that only occurs on one platform to be an indication of a bug across all platforms, but it "accidentally works" on the others because it lands within C++'s undefined behavior. So as long as we don't know why the segfault occurs, it's more important than just supporting pypy.