Still, that's not exception-safe, correct? If LogFlush or SleepForMicroseconds throws an exception the mutex will be unlocked twice, which pthreads disallows for normal mutexes...
You know, for a second I thought you were wrong, but I changed my mind. This does look like a bug, and a simple on to avoid at that.
It's tough, because Rocks is still highly based on LevelDB, which conforms to Google's coding style guideline, which makes RAII more than a bit tricky to do right.
It's tough, because Rocks is still highly based on LevelDB, which conforms to Google's coding style guideline, which makes RAII more than a bit tricky to do right.