Add MachMultiprocess and its test

Add MachMultiprocess and its test.

TEST=util_test MachMultiprocess.MachMultiprocess
R=rsesek@chromium.org

Committed: https://chromium.googlesource.com/crashpad/crashpad/+/bbe4d85935d7

[Mark Mentovai, 2014-08-16 06:39:37]

I think you’ll enjoy this one.

I also think that this is the last dependency before we can start doing the fun
process_types stuff.

[Robert Sesek, 2014-08-19 17:27:52]

Yes, this is cool. I probably could have used this in //sandbox/mac. Oh well :).

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
File util/test/mac/mach_multiprocess.cc (right):

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:98: if (pid > 0) {
Split this into private helper routines DoParent() and DoChild().

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:122: // Comb through the entire message,
checking every field against its
This might be excessive, but that's fine.

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:199: remote_port_ = MACH_PORT_NULL;
Do you need to do this? Won't the AutoReset do it for you? Or do you want it
done before calling waitpid()?

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:244: kr = bootstrap_look_up(bootstrap_port,
service_name.c_str(), &remote_port);
This doesn't require that stupid const_cast<>?

[Mark Mentovai, 2014-08-20 00:08:48]

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
File util/test/mac/mach_multiprocess.cc (right):

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:98: if (pid > 0) {
rsesek wrote:
> Split this into private helper routines DoParent() and DoChild().

I didn’t do this initially to keep the scopers properly scoped, but I’ve kept
the proper scoping while splitting it up with the aid of a helper struct.

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:199: remote_port_ = MACH_PORT_NULL;
rsesek wrote:
> Do you need to do this? Won't the AutoReset do it for you? Or do you want it
> done before calling waitpid()?

This wound up getting removed in the Run{Parent,Child}() refactor.

https://codereview.chromium.org/482483002/diff/60001/util/test/mac/mach_multi...
util/test/mac/mach_multiprocess.cc:244: kr = bootstrap_look_up(bootstrap_port,
service_name.c_str(), &remote_port);
rsesek wrote:
> This doesn't require that stupid const_cast<>?

It does for the 10.5 SDK, but I guess I’m OK with using the 10.6 SDK for 10.5
support for now.

I needed the const_cast for bootstrap_register() because that’s declared badly
in all SDKs.

[Robert Sesek, 2014-08-20 17:54:24]

LGTM

[Mark Mentovai, 2014-08-20 17:58:46]

Committed patchset #5 manually as bbe4d85935d7 (presubmit successful).