While testing a stack overflow handling change on OSX, I have discovered that when stack overflow happens in native runtime code (QCALL, FCALL, ...), we don't try to invoke the handler in the runtime since we consider the location unsafe. The result is that the process is terminated with SIGSEGV without any message.
I wonder if for stack overflow, we could be less strict in what location we would consider safe for handling this hardware exception and risk that in rare cases, we could hit a deadlock.
But we should print the stack overflow message again, at least.