Add ExceptionPorts and its test

Add ExceptionPorts and its test.

This is a unified interface to
{thread,task,host}_{get,set}_exception_ports().

TEST=util_test ExceptionPorts.*
R=rsesek@chromium.org

Committed: https://chromium.googlesource.com/crashpad/crashpad/+/8c8089dc7725f4cc843dd47625ce280234071911

[Mark Mentovai, 2014-09-15 19:06:41]

mark@chromium.org changed reviewers:
+ rsesek@chromium.org

[Mark Mentovai, 2014-09-15 19:06:42]

This depends on https://codereview.chromium.org/563383002/ for ExcMaskAll().

[Robert Sesek, 2014-09-16 17:38:09]

Need to review the test still, but also stepping away. Here's some comments.

https://codereview.chromium.org/549023005/diff/20001/util/mach/exception_port...
File util/mach/exception_ports.cc (right):

https://codereview.chromium.org/549023005/diff/20001/util/mach/exception_port...
util/mach/exception_ports.cc:69: MACH_LOG(ERROR, kr) << "mach_port_deallocate";
MACH_LOG_IF ?

https://codereview.chromium.org/549023005/diff/20001/util/mach/exception_port...
util/mach/exception_ports.cc:97: if (kr != KERN_SUCCESS) {
In many other cases, you return bool and then log the error for the caller; Get
and Set here behave differently.

https://codereview.chromium.org/549023005/diff/20001/util/mach/exception_ports.h
File util/mach/exception_ports.h (right):

https://codereview.chromium.org/549023005/diff/20001/util/mach/exception_port...
util/mach/exception_ports.h:96: //!     `*_get_exception_ports()` or
`*_set_exception_ports()`. The target
Does this pass ownership?

https://codereview.chromium.org/549023005/diff/20001/util/mach/exception_port...
util/mach/exception_ports.h:123: //! \param[in] port A send right to a Mach port
that will handle exceptions
Is the right consumed?

[Mark Mentovai, 2014-09-16 18:10:13]

Patchset #3 (id:40001) has been deleted

[Mark Mentovai, 2014-09-16 18:11:17]

Patchset #3 (id:60001) has been deleted

[Mark Mentovai, 2014-09-16 18:12:48]

Good suggestions, thanks. Updated.

[Robert Sesek, 2014-09-16 21:16:42]

One question: is the implementation/interface for setting the host ports just
for completeness, or do you expect any practical use?

https://codereview.chromium.org/549023005/diff/80001/util/mach/exception_port...
File util/mach/exception_ports_test.cc (right):

https://codereview.chromium.org/549023005/diff/80001/util/mach/exception_port...
util/mach/exception_ports_test.cc:186: if (code_count > 1) {
code_count has to be 2, or the test will fail. Not sure why this is branching.

[Mark Mentovai, 2014-09-16 21:29:11]

The host port stuff is actually usable from the command-line tool named
exception_port_tool, which you’ll wind up seeing soon. I do find it very useful
to be able to examine and change the host-level exception ports on a system for
debugging and testing purposes.

https://codereview.chromium.org/549023005/diff/80001/util/mach/exception_port...
File util/mach/exception_ports_test.cc (right):

https://codereview.chromium.org/549023005/diff/80001/util/mach/exception_port...
util/mach/exception_ports_test.cc:186: if (code_count > 1) {
rsesek wrote:
> code_count has to be 2, or the test will fail. Not sure why this is branching.

I have to guard it because I can’t use ASSERT_EQ in this non-void-returning
function. I added a comment explaining the problem. I don’t want to dereference
invalid memory and crash or test the wrong thing if code[0] is invalid. It’s
enough that the EXPECT_EQ failure will already cause the test to fail.

However, |> 1| was wrong, it should be |>=|. I fixed it here and in the other
locations where it was used.

[Mark Mentovai, 2014-09-16 21:29:24]

Patchset #4 (id:100001) has been deleted

[Robert Sesek, 2014-09-16 21:32:00]

LGTM

[Mark Mentovai, 2014-09-16 21:36:38]

Committed patchset #5 (id:140001) manually as 8c8089d (presubmit successful).