[SR-1701] Fatal errors in foundation should report backtraces

[phausler]

Previous ID	SR-1701
Radar	None
Original Reporter	@phausler
Type	Bug

Additional Detail from JIRA

Votes	0
Component/s	Foundation
Labels	Bug, StarterBug
Assignee	None
Priority	Medium

md5: ff5b9e13cf14794077788059c1f04d7d

Issue Description:

Most reasonable linux distributions have the function backtrace_symbols which can produce a sensible backtrace. It would be useful to emit failures with some call stack information (even if it was still mangled)

Likely this API would have to be added as a helper function in CoreFoundation due to it's funky syntax and requirements. But the call for this could encapsulate a fatal error and emit the debug information before terminating.

[ddunbar]

See SR-1700 for an example of where this would have been useful, particularly if we also include a little extra information in the fatalError().

[belkadan]

I think we really want backtraces everywhere. @gribozavr?

[phausler]

If at least for the Foundation assertions (we have a number of them) and perhaps if only in debug builds. This will aide in nailing down bugs versus expected behavior

[gribozavr]

We turn off backtraces in release builds because of security reasons. If the process has requested to stop, we should stop right away. We should not be trying to execute more code, which can increase the attack surface.