update blksize_t in raw to reflect libc change under openbsd

[semarie]

make os::raw::tests::unix test to pass again

r? @alexcrichton

[alexcrichton]

@bors: r+ b1d8554

[alexcrichton]

@bors: rollup

[aturon]

I just wanted to check in on this: we had mega problems making a similar change on Android, which was similarly considered having non-first tier support. Are we confident this won't instigate the same sort of breakage and dependency hell we saw there?

[semarie]

@aturon I am not sure to exactly understand the question. Some days ago, I updated OpenBSD Rust-libc based on libc-test (test suite of libc), and checking the C code. It results several code update to adjust Rust-libc with C-libc (see rust-lang/libc#110 for all changes).

blksize_t change (for OpenBSD) is part of these changes:

• switch to signed int32 in Rust-libc: semarie/rust-libc@b1a108c
• definition of __blksize_t (which is same as blksize_t) under C-libc: http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/sys/sys/_types.h?rev=1.9&content-type=text/x-cvsweb-markup

[alexcrichton]

@aturon yes I suspect this to have far less breakage for two reasons:

• The blksize_t type is much more rarely used than c_char
• OpenBSD is also more rarely used right now than even Android

I suspect it won't take long to mitigate any breakage that comes up, even if it does so.