Graham Markall

(partial) Python backtrace at segfault: ``` (gdb) py-bt Traceback (most recent call first): File "/home/gmarkall/numbadev/llvmlite/llvmlite/binding/ffi.py", line 151, in __call__ return self._cfn(*args, **kwargs) File "/home/gmarkall/numbadev/llvmlite/llvmlite/binding/targets.py", line 154, in get_pointee_abi_size size =...

The problem is that the context doesn't ensure that the model for a deferred type has been defined prior to trying to get its actual size. The following hack: ```diff...

It could be a duplicate or a separate issue - this maybe depends on whether we think defining a deferred type should also ensure that the data model is aware...

I just came across #6209 which is somewhat related to this.

This shouldn't be stale - removing "needtriage" so it doesn't go stale again.

Thanks for the report - when do you see this message? E.g. on import, or otherwise?

I'll re-check what OpenMP versions are available on currently-supported platforms, in case it's possible for us to just use `omp_set_max_active_levels` now.

As we support OSX back to 10.9, I'd guess this might not be simple to just fix by switching the deprecated method for the recommended replacement at the moment. @tamasgal...

Now I look at it, I'm not even sure why we call `omp_set_nested(0x1)`. According to the OpenMP 3.1 (July 2011, https://www.openmp.org/wp-content/uploads/OpenMP3.1.pdf p.30) and the 4.0 (July 2013, https://openmp.org/wp-content/uploads/OpenMP4.0.0.pdf p.36) specifications,...

Testing removal of this call in https://github.com/numba/numba/pull/7511.