add weak linkage to the ARM AEABI division functions

[Lokathor]

Closes #477

I know in the issue Amanieu said that all functions in this crate should probably get weak linkage when possible, but that seems overkill for my own needs and i didn't want to have to determine when the target will use ELF or not for a lot of functions i'd never heard of.

[Lokathor]

Note: It was mentioned in the issue that these functions are also incorrect on v4T. I'll resolve that as a separate PR so that we're just talking about one thing at a time.

[Lokathor]

I didn't touch those memory functions, so the test error seem unrelated.

[Lokathor]

@Amanieu ping

[Amanieu]

Sorry about the delay, I'm still looking into the powerpc issue. It seems like it might be an LLVM regression.

[Amanieu]

CI should be fixed now, could you rebase on the latest master?

[Amanieu]

Could you make the weak linkage conditional on !windows && !apple? Basically only on ELF targets.

[Lokathor]

do you detect for apple like this?

    #[cfg_attr(all(not(windows), not(target_vendor="apple")), linkage = "weak")]

or something else?

[Amanieu]

Yes, that's correct.

[Lokathor]

should be done i guess

[Lokathor]

also, does this automatically get picked up in every nightly release once merged, or is there a separate PR required for that?

[Amanieu]

I need to publish a new version of the crate (I want to get #474 merged first), and then a PR is needed in rust-lang/rust to update the crate version in Cargo.lock.

[Amanieu]

I've published version 0.1.77.