[Mac] Do not forward graceful shutdown from child signal handlers.

[Mac] Do not forward graceful shutdown from child signal handlers.

OSX implements fork() as a userland wrapper around a system call.  If
a graceful-shutdown signal is received before the graceful-shutdown
pipe has been replaced, the parent process will shutdown.  Change to
force a crash in the child in this case.

BUG=175341
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=184805

[Scott Hess - ex-Googler, 2013-02-25 20:33:31]

I had an alternate implementation which sent the pid in the message, but could
think of no added utility from doing so.

[Scott Hess - ex-Googler, 2013-02-25 20:55:01]

BTW, I tested by reproducing and seeing a backtrace in the appropriate place in
the child process.

[viettrungluu, 2013-02-25 23:15:06]

LGTM w/nits, but you break it, you buy it. You may also want to find a true
POSIX/Linux expert to double-check this too.

https://codereview.chromium.org/12314106/diff/1/chrome/browser/chrome_browser...
File chrome/browser/chrome_browser_main_posix.cc (right):

https://codereview.chromium.org/12314106/diff/1/chrome/browser/chrome_browser...
chrome/browser/chrome_browser_main_posix.cc:47: // which registered
|g_shutdown_pipe_write_fd|.
Could you include a reference to the bug(s)?

https://codereview.chromium.org/12314106/diff/1/chrome/browser/chrome_browser...
chrome/browser/chrome_browser_main_posix.cc:311: g_pipe_pid = getpid();
Nit: You may as well initialize these in the same order as they're declared
(i.e., g_pipe_pid before the fds).

[Scott Hess - ex-Googler, 2013-02-25 23:20:45]

Conveniently, an OWNERS line corresponds to a Linux expert.

getpid() is async signal safe, in case that comes up.

https://codereview.chromium.org/12314106/diff/1/chrome/browser/chrome_browser...
File chrome/browser/chrome_browser_main_posix.cc (right):

https://codereview.chromium.org/12314106/diff/1/chrome/browser/chrome_browser...
chrome/browser/chrome_browser_main_posix.cc:47: // which registered
|g_shutdown_pipe_write_fd|.
On 2013/02/25 23:15:06, viettrungluu wrote:
> Could you include a reference to the bug(s)?

Done.

https://codereview.chromium.org/12314106/diff/1/chrome/browser/chrome_browser...
chrome/browser/chrome_browser_main_posix.cc:311: g_pipe_pid = getpid();
On 2013/02/25 23:15:06, viettrungluu wrote:
> Nit: You may as well initialize these in the same order as they're declared
> (i.e., g_pipe_pid before the fds).

Done.

I was not willing to initialize the write fd before the read fd.  It looked too
weird.  Should have used a socketpair(), or just assumed that pipe() _is_ a
socketpair(), but  that would be creepy, too.

[Scott Hess - ex-Googler, 2013-02-26 18:22:49]

Err, Lei, should have made more clear that the OWNERS callout was for you.

[Lei Zhang, 2013-02-26 21:41:22]

lgtm

[commit-bot: I haz the power, 2013-02-26 21:45:23]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/shess@chromium.org/12314106/7001

[Scott Hess - ex-Googler, 2013-02-27 00:53:59]

Committed patchset #3 manually as r184805 (presubmit successful).

[Scott Hess - ex-Googler, 2013-02-27 00:57:40]

On 2013/02/27 00:53:59, shess wrote:
> Committed patchset #3 manually as r184805 (presubmit successful).

Apologies if those redness were actually real.  Looking at the logs, I find it
impossible to tell.