Emergency Self-Construct

Results 879 comments of Emergency Self-Construct

I am not certain why the CI is failing. Initially I assumed that this was unrelated to this PR and that `main` was broken. But I ran that here: https://dev.azure.com/numba/numba/_build/results?buildId=12723&view=results...

> I suspect the cause is my change to the float unboxing. I also suspect the failing test is one that I wrote a while ago! I see, it's this...

> I was planning to come back and debug this in the next week or so. Sounds good, FYI: the next RC is still some months away.

@eric-wieser I see you have moved the code to `cgutils`. Sorry for not being more clear here, I think the code should remain in `boxing.py` but ought to be re-exported...

@eric-wieser thank you, I have made both changes!

@eric-wieser you can merge in `main` to fix the RTD build failure

> > We simplified it from a function where the literal_unroll actually did something useful, this was just the shortest example we could find that still resulted in a segfault....

@twiecki @aseyboldt I can confirm I am able to reproduce, but only using packages from `conda-forge`, not using those from the `numba` channel. If you setup an environment on M1...

> @esc My above example still segfaults for me even when installing numba from the numba channel. I still can't reproduce this, did you try with a completely fresh environment?