support no_std

[dragazo]

I have a compiler/interpreter project I'm developing for embedded systems, and this crate looks perfect for saving a ton of heap space and gaining locality for my many symbol tables. The only issue is that it needs to be no_std and #29 was closed a long time ago. So I've gone ahead and redone that work here.

It's pretty much all just a switch from std to core or alloc versions of things, plus a new (default) feature flag called std that enables a few small things like operations on OsString and the like.

The new version passes all test cases in standard mode, successfully builds in no-std mode, and miri has no issues with anything.

[ParkMyCar]

Thanks for the PR @dragazo! One of the reasons I closed #29 was I couldn't figure out how to test for no_std support in CI. I would want to make sure we have an automated way to detect regressions here. Do you think you could try adding this?

P.S. your compiler/interpreter project sounds very cool!

[dragazo]

Oh sure, I have a trick I use to test that in my no-std libs, which is to cross-compile targeting thumbv6m-none-eabi, which does not (and likely will never) support the standard library. I just added a new github workflow for that.

The only issue it discovered was caused by a missing LifetimeFree impl for String, which is a no-std compatability issue in castaway. I've already fixed that in a fork, which is pending approval here. Once that's merged and eventually published, we'll just need to switch out to the new alloc feature I added for that dep, and it should pass the CI I just added to this repo.

[ParkMyCar]

Oh sure, I have a trick I use to test that in my no-std libs, which is to cross-compile targeting thumbv6m-none-eabi, which does not (and likely will never) support the standard library. I just added a new github workflow for that.

Neat! That's a good trick to know

The only issue it discovered was caused by a missing LifetimeFree impl for String, which is a no-std compatability issue in castaway. I've already fixed that in a fork, which is pending approval here. Once that's merged and eventually published, we'll just need to switch out to the new alloc feature I added for that dep, and it should pass the CI I just added to this repo.

Sounds great, thanks for putting up a PR against castaway!