The regression tests in this directory are all concerned with testing
whether libmoopriv successfully managed to wrap everything it needs to.

Each .c file is built into a separate binary. Each binary and perl script
is run twice, once in a normal unix environment, and once with libmoopriv
preloaded. If it outputs anything different to stdout, or returns a
different return code with libmoopriv preloaded, the test failed. Moreover,
the run with libmoopriv preloaded must output a line starting with "mooix:"
to stderr. That line must contain the first thing the test script outputs to
stdout. This is to verify that the moo wrapper did wrap the given function.

The remainder of stderr must be identical with what the first run outputs, 
if anything.

This can't be easily used to test calls that actually need to go through
mood, since mood can't be assumed to be running for the regression tests.
So the tests take care to not trigger the proxy to try to talk to mood.
Generally that's doable, but it's right out for anything involving forking,
unfortuntatly.
