Emergency Self-Construct

@Faberix thank you, I can confirm that I observe this not halting too. My guess is that it's a bug in the type inference algorithm. You can use `NUMBA_DEBUG=1` to...

@larsoner thank you for bringing this to our attention. This may be a `pip` bug, I found this: https://github.com/pypa/pip/issues/11835

> Honestly, I don't think Numba should write into this directory at all. It has no business adding cache files in a place like this. @hoechenberger sorry to hear that...

FWIW: these are the docs about how Numba will select the cache: https://numba.readthedocs.io/en/stable/reference/envvars.html#envvar-NUMBA_CACHE_DIR Perhaps this order could be made configurable and so libraries that depend on Numba could decide to...

@hoechenberger @larsoner thank you for your feedback! > > So, if we were to set this env var in our package before importing numba, the setting will be respected? >...

> I'm not sure how conda would behave in this scenario, this is something to be considered as this tool is widely used in scientific Python package management. Would be...

@hoechenberger great, thank you for checking. It is my guess that a more fine grained control over Numba's caching mechanism will likely help alleviate some of the pain for library...

@larsoner @hoechenberger -- thank you again for looking into this. I have created a feature request based on our conversation in this bug report: https://github.com/numba/numba/issues/9350 Please do comment in case...

> One thing I think is missing is the "bug fix" idea, namely that if you do put files in the installed directory then you should modify the RECORD as...

As discussed in: https://github.com/numba/numba/pull/9309#discussion_r1409148087 the PR to implement this should also fixup: * [ ] The template to have the correct double and single underline. * [ ] The existing...