Enough to often go nuts working with it but not enough that it rolls off the finger tips.
I actually tried swapping out the Git submodule version of RocksDB which is rocksdb @ bd68bfb for a current one and ran into a new set of errors.
The one requested by the RocksDB pack and the commits of RocksDB are compiling clean by themselves but when built via pack_install/1 or pack_rebuild/1 lead to errors. I tried running make shared_object in the pack directory but that is failing because pack_install/1 is setting up some environment variables that are not there when make is executed directly.
Saw this post that notes
For debugging purposes pack_install creates a shell script in the build location that sets the same variables, so you can run the compiler from any environment.
but I can not find the script or even in the Prolog code where it is created. Needless to say this is my typical scenario with make and C++, try, back step try again, and again and again and then one day wake up lucky and it just works. Then it seems I can never get such things to fail again.