Make check doesn't prevent people from breaking make check-stage0-std

[bblum]

When developing inside of libstd, it's a lot more convenient to be able to make check-stage0-std instead of make check-stage2-std. But every now and again, we lose the ability to do that to some obscure linker error which can languish for a long time because the test suite doesn't make sure that it works.

Currently it is broken due to this:

error: duplicate entry for `str_eq`
error: aborting due to previous error

[brson]

The current breakage is because of this declaration:

/// Bytewise slice equality                                                                                                                                                      
#[cfg(test, stage0)]
#[lang="str_eq"]
#[inline]
pub fn eq_slice(a: &str, b: &str) -> bool {
    do a.as_imm_buf |ap, alen| {
        do b.as_imm_buf |bp, blen| {
            if (alen != blen) { false }
            else {
                unsafe {
                    libc::memcmp(ap as *libc::c_void,
                                 bp as *libc::c_void,
                                 (alen - 1) as libc::size_t) == 0
                }
            }
        }
    }
}

It shouldn't be declared as a lang item - it is explicitly creating a duplicate lang item for stage0 stdtest

[alexcrichton]

I don't think that this is really an issue that's completely actionable. The best takeaway from this is "make snapshots more often". There are very real reasons I think why make check-stage0-std should break sometimes (while stage0 changes are being made).

Hopefully the windows in which make check-stage0-std is broken are fairly small, but I think that there's not much to do on this.