Investigate Native SDK Backend/Session behaviour on Shutdown on M1

[ashwoods]

We would want to verify how the Native SDK behaves in these instances:

Various ways of process termination:

• KILL / TERM
• Logout / Shutdown

Using the following backends:

• crashpad
• breakpad

All that, specifically on macOS on the M1.

STR:

Possibly patch up the native example, increasing this sleep here:

sentry-native/examples/example.c

Line 177 in bd56100

sleep_s(10);

Then build the example with the various backends, and start it like so:
sentry_example log start-session sleep.
… and then kill it in various ways.
Observe if a crash report and which kind of session status is being reported.

[supervacuus]

I can do the testing and create further issues if I see any unwanted behavior.

[Swatinem]

The specific case we would like to confirm in our testing, which came from a customer:
They report that with crashpad, when the process is forcibly killed on shutdown due to taking too long to terminate, they observe a crash log, which in turn flags the session as crashed due to filesystem timestamps.

Not quite sure this will necessarily exhibit that case, maybe you would have to install a handler for SIGTERM or whatever is signalling the process to pls shut down; and then run into a timeout there.

The desired outcome here would be: don’t create a crashed session if the process is being forcibly killed on shutdown, but rather a abnormal session.

[supervacuus]

@ashwoods, @Swatinem, since this was coming from support: was this ever raised again? I would like to close this otherwise.

[ashwoods]

Hasn't resurfaced since, closing now.