Linux: move SyncIPC function from Skia to base.

Linux: move SyncIPC function from Skia to base.

I'll be needing in some pending WebKit changes.

[Evan Martin, 2009-06-18 00:36:07]

http://codereview.chromium.org/131006/diff/1/2
File base/unix_domain_socket_posix.cc (right):

http://codereview.chromium.org/131006/diff/1/2#newcode100
Line 100: int fds[2];
2spc tabs

comment in here that this socketpair is used just for this single send/recv, and
cleaned up before returning

http://codereview.chromium.org/131006/diff/1/2#newcode114
Line 114: const ssize_t r = RecvMsg(fds[0], reply, reply_len, &fd_vector);
would be nice to not use a single letter variable here, especially when it's
used 21 lines later

http://codereview.chromium.org/131006/diff/1/2#newcode119
Line 119: if ((fd_vector.size() > 0 && result_fd == NULL) || fd_vector.size() >
1) {
It seems these sorts of errors are not like the SendMsg one, in that they're
indicating something has more or less gone wrong at compile-time.  That is, if
you're getting a reply it shouldn't have its vector messed up.  Would it make
sense to NOTREACHED here?

http://codereview.chromium.org/131006/diff/1/3
File base/unix_domain_socket_posix.h (right):

http://codereview.chromium.org/131006/diff/1/3#newcode10
Line 10: 
This file is getting bigger, can you put a description at the top?  ("sync IPCs
for blah blah blah, see this wiki page")

http://codereview.chromium.org/131006/diff/1/3#newcode23
Line 23: // along with a file descriptor over which the reply is read.
The way this is worded, it sounds like you're sending a file descriptor, which I
don't think you are.