Jay R. Ashworth wrote:
On Mon, Jan 15, 2007 at 11:45:13AM -0500, Simetrical
On 1/15/07, Edward Z. Yang
WEll, it totally depends on the culture
you've made regarding unit
tests. Here, it's "Oh, it's unreasonable to think that all the
parsertests will pass" but in other places its "Unit tests are super
important and no development goes forward until all tests pass". I
myself fall into the latter camp.
In case you didn't notice, though, 18 of the
19 failing tests are
actually unimplemented features, not regressions. :) Maybe it would
be best if we removed the TODOs altogether. Then we would get,
appropriately, passes when we have no regressions, and failures on
regressions. People might actually even notice when regressions crop
up! (How long has that 1887 part 2 thing been regressed?)
I concur that this seems an excellent idea. If the terror level, er,
um, error level is always greater than 0 then no one will pay any
attention to it.
There could still be a separate list of parser tests we want to pass.