linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

StackDumpSignalHandler is used by EnableInProcessStackDumping, usually on non-
official builds. On Mac it prints the stack and re-raises the signal. On Linux
it prints the stack and calls _exit(1). This means that Linux and Chrome OS
builds won't leave core files or call crash_reporter. This is super-confusing
for developers, since most of us don't run official builds.

Print a warning that the crashing process is calling _exit(1) and that a core
file will not be generated.

BUG=551681
TEST=manual, trigger a CHECK, see the warning printed

Review-Url: https://codereview.chromium.org/2694023002
Cr-Commit-Position: refs/heads/master@{#450192}
Committed: https://chromium.googlesource.com/chromium/src/+/bb9edce13cdf6f840b3a940d3649d5631c660544

[James Cook, 2017-02-13 19:27:39]

The CQ bit was checked by jamescook@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-02-13 19:28:24]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-02-13 20:03:58]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-02-13 20:03:59]

Dry run: Try jobs failed on following builders:
  cast_shell_linux on master.tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/cast_shell_linu...)

[James Cook, 2017-02-13 20:15:16]

Description was changed from

==========
WIP

BUG=
==========

to

==========
Re-raise signal in StackDumpSignalHandler (used by EnableInProcessStackDumping)

Right now a crash on a non-official-build of Linux/Chrome OS will print a stack
trace and then _exit(1). This means that on-device Chrome OS builds won't leave
core files. It also makes it confusing to work on crash reporting.

BUG=551681
==========

[James Cook, 2017-02-13 21:58:05]

Description was changed from

==========
Re-raise signal in StackDumpSignalHandler (used by EnableInProcessStackDumping)

Right now a crash on a non-official-build of Linux/Chrome OS will print a stack
trace and then _exit(1). This means that on-device Chrome OS builds won't leave
core files. It also makes it confusing to work on crash reporting.

BUG=551681
==========

to

==========
linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

StackDumpSignalHandler is used by EnableInProcessStackDumping, usually on non-
official builds. On Mac it prints the stack and re-raises the signal. On Linux
it prints the stack and calls _exit(1). This means that Linux and Chrome OS
builds won't leave core files or call crash_reporter. This is super-confusing
for developers, since most of us don't run official builds.

Print a warning that the crashing process is calling _exit(1) and that a core
file will not be generated.

BUG=551681
TEST=manual, trigger a CHECK, see the warning printed
==========

[James Cook, 2017-02-13 21:58:28]

The CQ bit was checked by jamescook@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-02-13 21:59:24]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[James Cook, 2017-02-13 22:02:32]

Description was changed from

==========
linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

StackDumpSignalHandler is used by EnableInProcessStackDumping, usually on non-
official builds. On Mac it prints the stack and re-raises the signal. On Linux
it prints the stack and calls _exit(1). This means that Linux and Chrome OS
builds won't leave core files or call crash_reporter. This is super-confusing
for developers, since most of us don't run official builds.

Print a warning that the crashing process is calling _exit(1) and that a core
file will not be generated.

BUG=551681
TEST=manual, trigger a CHECK, see the warning printed
==========

to

==========
linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

StackDumpSignalHandler is used by EnableInProcessStackDumping, usually on non-
official builds. On Mac it prints the stack and re-raises the signal. On Linux
it prints the stack and calls _exit(1). This means that Linux and Chrome OS
builds won't leave core files or call crash_reporter. This is super-confusing
for developers, since most of us don't run official builds.

Print a warning that the crashing process is calling _exit(1) and that a core
file will not be generated.

BUG=551681
TEST=manual, trigger a CHECK, see the warning printed
==========

[James Cook, 2017-02-13 23:16:18]

jamescook@chromium.org changed reviewers:
+ erikchen@chromium.org

[James Cook, 2017-02-13 23:16:18]

erikchen, please take a look. I hate to add logging, but I've been burned twice
by this, for a total of 2 man-days of wasted time, while working on crash
reporting for mustash. I tried to fix the underlying issue and make it re-raise
signals, but I didn't make any headway on the underlying Chrome OS test
failures.

[erikchen, 2017-02-13 23:36:30]

On 2017/02/13 23:16:18, James Cook wrote:
> erikchen, please take a look. I hate to add logging, but I've been burned
twice
> by this, for a total of 2 man-days of wasted time, while working on crash
> reporting for mustash. I tried to fix the underlying issue and make it
re-raise
> signals, but I didn't make any headway on the underlying Chrome OS test
> failures.

lgtm.

I believe I added the Mac reraising for the same reason [crash reporting doesn't
work! https://bugs.chromium.org/p/chromium/issues/detail?id=549379]. I
personally think that StackDumpSignalHandler() shouldn't exist [different
behavior on official builds, and for exactly the reason you just ran into], but
I didn't feel like making a stand.

[James Cook, 2017-02-13 23:55:12]

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

[James Cook, 2017-02-13 23:55:13]

mark, can I get OWNERS?

As I mentioned to Erik above, I hate to add logging, but I've been burned twice
by StackDumpSignalHandler eating signals, for a total of 2 man-days of wasted
time, while working on crash reporting for mustash. I tried to fix the
underlying issue and make it
re-raise signals, but I didn't make any headway on the underlying Chrome OS test
failures.

[commit-bot: I haz the power, 2017-02-13 23:57:47]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-02-13 23:57:48]

Dry run: This issue passed the CQ dry run.

[Mark Mentovai, 2017-02-14 01:18:23]

LGTM because this at least makes what’s happening obvious, but
StackDumpSignalHandler is a mess. Also, getting re-raise right can be tricky.

[James Cook, 2017-02-14 01:29:30]

The CQ bit was checked by jamescook@chromium.org

[commit-bot: I haz the power, 2017-02-14 01:30:11]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-02-14 01:38:14]

CQ is committing da patch.

Bot data: {"patchset_id": 20001, "attempt_start_ts": 1487035770869960,
"parent_rev": "ed9773f95712d0477ef7f3f2e7be3d6d79d934b1", "commit_rev":
"bb9edce13cdf6f840b3a940d3649d5631c660544"}

[commit-bot: I haz the power, 2017-02-14 01:39:06]

Description was changed from

==========
linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

StackDumpSignalHandler is used by EnableInProcessStackDumping, usually on non-
official builds. On Mac it prints the stack and re-raises the signal. On Linux
it prints the stack and calls _exit(1). This means that Linux and Chrome OS
builds won't leave core files or call crash_reporter. This is super-confusing
for developers, since most of us don't run official builds.

Print a warning that the crashing process is calling _exit(1) and that a core
file will not be generated.

BUG=551681
TEST=manual, trigger a CHECK, see the warning printed
==========

to

==========
linux/cros: Print a warning when StackDumpSignalHandler calls _exit(1)

StackDumpSignalHandler is used by EnableInProcessStackDumping, usually on non-
official builds. On Mac it prints the stack and re-raises the signal. On Linux
it prints the stack and calls _exit(1). This means that Linux and Chrome OS
builds won't leave core files or call crash_reporter. This is super-confusing
for developers, since most of us don't run official builds.

Print a warning that the crashing process is calling _exit(1) and that a core
file will not be generated.

BUG=551681
TEST=manual, trigger a CHECK, see the warning printed

Review-Url: https://codereview.chromium.org/2694023002
Cr-Commit-Position: refs/heads/master@{#450192}
Committed:
https://chromium.googlesource.com/chromium/src/+/bb9edce13cdf6f840b3a940d3649...
==========

[commit-bot: I haz the power, 2017-02-14 01:39:07]

Committed patchset #2 (id:20001) as
https://chromium.googlesource.com/chromium/src/+/bb9edce13cdf6f840b3a940d3649...