Collect the cflow data for FFMpeg. Use the --reverse option to get the most nodes we can, knowing that mostly will be disconnected. Then, look into their code coverage tools to determine if there's a way we can get runtime monitoring on their automated test suite. What I'm looking for is something like a "get a backtrace every 5ms" thing that we can use to determine a more complete call graph. A coverage tool probably doesn't do THAT, but a profiling tool does.
Collect the cflow data for FFMpeg. Use the --reverse option to get the most nodes we can, knowing that mostly will be disconnected. Then, look into their code coverage tools to determine if there's a way we can get runtime monitoring on their automated test suite. What I'm looking for is something like a "get a backtrace every 5ms" thing that we can use to determine a more complete call graph. A coverage tool probably doesn't do THAT, but a profiling tool does.
The key here is to figure out a tool set.