On Mar 9, 2009, at 11:48 AM, Aurélio A. Heckert wrote:

May we need to get seriously in a test-driven development.

Running the "make check" now we see "Failed 16 of 176 tests".

If we have more tests and if we test before commit, the regression

is less possible.

How about a TDD politic?

Probably about time.

I was one who set up a lot of the unit testing (though not the first). It's been a while, but we probably need to buckle down and finish getting those last 16 tests passing.

As soon as we get them set, we can enforce the "no fails" restriction.

I seem to recall that at least one of the non-passing issues exposed a problem with how we were round-tripping some data, and might need something else to fix... but then again that's why we need tests.