Yeb Havinga <yebhavinga@gmail.com> writes:
> Anyway I now have troubles to make a good test case to trip the lbound
> check.
I don't believe you ever did have a test case that would trip the lbound
check. What was failing was the ARR_NDIM == 1 check. If you were to
look at the array in gdb, the apparent value of the lbound would be
indeterminate because it's off the end of the allocated space for a
zero-D array. But the code wouldn't be reaching that test.
regards, tom lane