0.1.178->0.1.179 regression: 32-bit musl `struct stat` fields

[tgross35]

It looks like 793fa97 removed the st_atime + st_atime_nsec and similar fields, which should still exist under #[cfg(not(musl_v1_2_3))]. Result:

https://docs.rs/libc/0.2.179/i686-unknown-linux-musl/libc/struct.stat.html
vs.
https://docs.rs/libc/0.2.178/i686-unknown-linux-musl/libc/struct.stat.html.

Cc @xbjfk

[xbjfk]

I'm sorry about this - this was a poor oversight on my part - I've created #4914 to fix this

[tgross35]

Not at all a problem, I should have caught it as well :) thank you for the quick fix!

[tgross35]

I'm actually not quite sure why tests didn't pick this up. There are some overrides

libc/libc-test/build.rs

Lines 3974 to 3980 in 17adf2d

	// Our stat *_nsec fields normally don't actually exist but are part
	// of a timeval struct - this is fixed in musl_v1_2_3
	("stat" | "statfs" | "statvfs" | "stat64" | "statfs64" | "statvfs64", f)
	if !musl_v1_2_3 && f.ends_with("_nsec") =>
	{
	Some(f.replace("e_nsec", ".tv_nsec"))
	}

but it should be getting checked.