As of right now, Tcl has some failing tests skipped, some todo'd, and
other simply failing. Any failing tests should be properly todo'd --
and the test harness should have some way of showing those as
failures. Then users can know that everything works and developers
can know what to fix.
APL does this with an environment variable. It sounds like a good way
to go. Ideally, there would be multiple make targets for testing so
developers or users could explicitly test the different conditions
without the use of an environment variable.
--
Matt Diephouse