[Cryptech Tech] More tests and a bug

[Randy Bush]

> I've yet to implement a way to programmatically check the results from
> the tests.

in the '80s, i was poisoned by the methods of 

   "PASCAL Compiler Validation", by B.A. Wichmann and Z.J. Ciechanowicz.

regression testing using a constantly enhanced suite of tests, each one
being a setup and run, collecting the output, diffing, and reporting
only if the diff shows change from the authoritative result.

we used it for pascal, modula-2, c, and c++ compiler testing.  yes, we
washed the machine after it touched any c++ :)

this is pretty standard and stupid stuff, which i consider a feature,
not a bug.  it does require building a framework so that each test is
fairly simple/small, and a test harness which can drive the device under
test.  [ i have this horrible feeling you are gonna like expect ]

randy