Add shadow password bindings: setspent, endspent, getspent, getspnam plus spwd, the corresponding struct

couldn't get strftime to work, even with help from IRC. Removing it now as it's not worth the hassle

Adding accept4 function

accept4 is a version of accept that takes flags.
It is more efficient for certain use cases, and has
a more modern path inside the kernel.

Adding POSIX 'E' error code constants for Windows

Believe it or not, Windows actually defines a few POSIX
'E' error codes (eg EINVAL). By adding them to libc, the Windows
bindings for things like `read()` become useful.

Believe it or not, Windows actually defines a few POSIX
'E' error codes (eg EINVAL). By adding them to libc, the Windows
bindings for things like `read()` become useful.

accept4 is a version of accept that takes flags.
It is more efficient for certain use cases, and has
a more modern path inside the kernel.

aarch64: test against Ubuntu 16.04

also "backport" two glibc fixes:

- ipc_perm.mode is c_uint. [commit][1] [discussion][2]
- bump SIGSTKSZ to 16384. [commit][3] [discussion][4]

so libc-test passes.

[1]: https://sourceware.org/git/?p=glibc.git;a=commit;h=5c40c3bab2
[2]: https://www.sourceware.org/ml/libc-alpha/2014-10/msg00446.html
[3]: https://sourceware.org/git/?p=glibc.git;a=commit;h=b763f6ae859ecea70a5dacb8ad45c71d5f667e2e
[4]: https://sourceware.org/ml/libc-alpha/2015-04/msg00033.html

closes #231

r? @alexcrichton

libc-test: cargo update

to pick up the recent changes in the ctest repo.

r? @alexcrichton

also "backport" two glibc fixes:

- ipc_perm.mode is c_uint. [commit][1] [discussion][2]
- bump SIGSTKSZ to 16384. [commit][3] [discussion][4]

so libc-test passes.

[1]: https://sourceware.org/git/?p=glibc.git;a=commit;h=5c40c3bab2
[2]: https://www.sourceware.org/ml/libc-alpha/2014-10/msg00446.html
[3]: https://sourceware.org/git/?p=glibc.git;a=commit;h=b763f6ae859ecea70a5dacb8ad45c71d5f667e2e
[4]: https://sourceware.org/ml/libc-alpha/2015-04/msg00033.html

closes #231

Factor common code out of the bitrig and openbsd modules

The new directory structure better reflects the historical relationship
of the BSDs, and includes 400 fewer duplicative lines of code.

The new directory structure better reflects the historical relationship
of the BSDs, and includes 400 fewer duplicative lines of code.

Diff reduction between OpenBSD and Bitrig

Bitrig had a number of errors and omissions.
Also, fill out struct glob_t on all BSDs.

A follow-up commit will condense OpenBSD/Bitrig common code into a separate file.

use new openbsd-6.0 qcow2 image

this version is using `openbsd-6.0-without-pkgs` image.

- allow QEMU name to have subdirectory inside (replace `/` by `__` in
  the filename)
- add a new code path for plain qcow2 image (for efficient bandwidth
  usage, the qcow2 is expected to use compressed qcow2 format)
- move freebsd image to explicitly download the gzipped QEMU

http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/sys/sys/sysctl.h?rev=1.162&content-type=text/x-cvsweb-markup
http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/sys/sys/sysctl.h.diff?r1=1.161&r2=1.162

Bitrig had a number of errors and omissions.
Also, fill out struct glob_t on all BSDs.

Bump to 0.2.16

Move kqueue-related definitions from nix to libc

This PR moves, as much as possible, kqueue and kevent related definitions from the nix crate to libc.  There will be a companion PR to the nix repo.

Add more s390x definitions

This adds more Linux-specific definitions for s390x, to match the
defines provided for other Linux targets.  Many of these will be
required to build std, rustc, and cargo.

In addition, this changes the type of sigaction.sa_flags to "int"
to match current glibc after this bug fix was applied:
https://sourceware.org/bugzilla/show_bug.cgi?id=16713



Some other (cosmetic) changes to bring types in line with the
exact types used in the glibc headers (e.g. ino64_t instead
of ino_t in statfs64).

Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>

Add support for sync_file_range on Linux

This adds more Linux-specific definitions for s390x, to match the
defines provided for other Linux targets.  Many of these will be
required to build std, rustc, and cargo.

In addition, this changes the type of sigaction.sa_flags to "int"
to match current glibc after this bug fix was applied:
https://sourceware.org/bugzilla/show_bug.cgi?id=16713



Some other (cosmetic) changes to bring types in line with the
exact types used in the glibc headers (e.g. ino64_t instead
of ino_t in statfs64).

Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>

* Rollback FreeBSD definitions to circa 10.3-RELEASE
* Remove EVFILT_SYSCOUNT across the board.  It's value is changeable, so
  it's not safe to define with FFI.

Add prelimenary wasm32 support

For the moment there's nothing different about the emscripten libc implementations for asmjs and wasm32.

mips: add SO_REUSEPORT

Which is required to cross compile the net2 crate. Which, in turn, is required to cross compile Cargo.

r? @alexcrichton

add support for s390x

tested by running libc-test on real s390x hardware.

r? @alexcrichton