Any specific instructions on reproducing the problem would be appreciated. Especially along the lines of
1. fire up this AMI on Amazon Web Services
2. Do this and that
Setting up non-default development platforms is a chore and having a reproducible error makes things easier.
I don't have AMI. Sorry. So basically I wrote a simple multi-thread application that just writes to this logger (I was regression testing it) where each of the thread (2 threads) writes to the same logger. I was using the logstream (I really prefer the C++ syntax of ostream) as you can probably guess from the stack.
Now the odd thing is, for the cases of when it throws exceptions, the crash happens in the beginning. And for the case where it SIGSEGV, it happens near the end as the ~logstream_base is called.
You'll see the differences when you see the two stack traces I provided. Did I cover all your questions? If not, ask away.