linux/android: add SIGTRAP to the list of signals handled by the client

linux/android: add SIGTRAP to the list of signals handled by the client

__builtin_trap() causes a SIGTRAP on arm64 (at least with GCC 4.9).
SIGTRAP is not handled by breakpad, causing crashes induced by
__builtin_trap() to be missed.
Note that on x86 and arm, instead, __builtin_trap() raises a SIGILL,
which is already handled by breakapd.

BUG=chromium:614865
R=vapier@chromium.org

Committed: https://chromium.googlesource.com/breakpad/breakpad/+/a34c929f6f46c1c414147ecfa56c3562b2eca027

[Primiano Tucci (use gerrit), 2016-06-06 18:16:32]

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

[Primiano Tucci (use gerrit), 2016-06-06 18:16:33]

So, I am extremely confident that this CL is going to backfire in my hands some
months from now. But right now I don't see any reason why we shouldn't be
handling SIGTRAP.
I checked and gdb still works after this CL. My understanding is that gdb has
some special routing of signals via the ptrace() interface and does not rely on
sigaction() signal handlers.

[Primiano Tucci (use gerrit), 2016-06-06 19:16:38]

On 2016/06/06 18:16:33, Primiano Tucci wrote:
> So, I am extremely confident that this CL is going to backfire in my hands
some
> months from now. But right now I don't see any reason why we shouldn't be
> handling SIGTRAP.
> I checked and gdb still works after this CL. My understanding is that gdb has
> some special routing of signals via the ptrace() interface and does not rely
on
> sigaction() signal handlers.

As suggested by rsesek@ I checked that also breakpoints and watches work both on
Linux desktop and Android after passing --enable-crash-reporter-for-testing.

[Primiano Tucci (use gerrit), 2016-06-08 15:52:48]

primiano@chromium.org changed reviewers:
+ thestig@chromium.org

[Primiano Tucci (use gerrit), 2016-06-08 15:52:49]

+thestig, mark seems OOO (I'm disappointed by the lack of an actual emoji,
buuuu)

[vapier, 2016-06-08 16:02:48]

vapier@chromium.org changed reviewers:
+ vapier@chromium.org

[vapier, 2016-06-08 16:02:48]

lgtm

[Primiano Tucci (use gerrit), 2016-06-08 16:04:53]

Description was changed from

==========
linux/android: add SIGTRAP to the list of signals handled by the client

__builtin_trap() causes a SIGTRAP on arm64 (at least with GCC 4.9).
SIGTRAP is not handled by breakpad, causing crashes induced by
__builtin_trap() to be missed.
Note that on x86 and arm, instead, __builtin_trap() raises a SIGILL,
which is already handled by breakapd.

BUG=chromium:614865
==========

to

==========
linux/android: add SIGTRAP to the list of signals handled by the client

__builtin_trap() causes a SIGTRAP on arm64 (at least with GCC 4.9).
SIGTRAP is not handled by breakpad, causing crashes induced by
__builtin_trap() to be missed.
Note that on x86 and arm, instead, __builtin_trap() raises a SIGILL,
which is already handled by breakapd.

BUG=chromium:614865
R=vapier@chromium.org

Committed:
https://chromium.googlesource.com/breakpad/breakpad/+/a34c929f6f46c1c414147ec...
==========

[Primiano Tucci (use gerrit), 2016-06-08 16:04:55]

Committed patchset #1 (id:1) manually as
a34c929f6f46c1c414147ecfa56c3562b2eca027 (presubmit successful).

[Mark Mentovai, 2016-06-09 16:48:18]

I had some great emoji lined up, but Rietveld’s only 1970s-era 7-bit-clean.

[Primiano Tucci (use gerrit), 2016-06-09 17:33:45]

On 2016/06/09 16:48:18, Mark Mentovai (emoji for OOO) wrote:
> I had some great emoji lined up, but Rietveld’s only 1970s-era 7-bit-clean.
Maybe that's why it still works :)

[Mark Mentovai, 2016-06-09 17:37:12]

I lied, it doesn’t accept esoteric characters like ! and & either. It’s a wonder
that it lets us use mixed case at all.