On 03 Jun 2014, at 15:07, Boris Zbarsky <bzbar...@mit.edu> wrote: > On 6/3/14, 8:50 AM, Boris Zbarsky wrote: >> I do think we should be very intentional about adopting something new, >> both in terms of semantics (mochitest is() using == is a mistake we >> should not duplicate in the short-name comparison function in the new >> setup) y > > One other note. > > The checkin so far preserved the assertion semantics in xpcshell, afaict: > failure throws and terminates the test.
Yes, deliberately so. > > I assume that the mochitest version will use a different reporter that > doesn't throw-and-terminate, to preserve the current semantics of mochitest > assertions. (If this assumption is incorrect, we need to have a separate > discussion about that.) If so, we'll have the same methods but different > semantics in the different harnesses. Not sure how much of a problem that is > in practice... not least because I don't think people actually write xpcshell > tests all that much. Ideally, run-until-failure would be a configurable option per test suite. For example, in automation we’d continue on failure and locally we’d stop on failure. Additionally, that’d be configurable per test and per run. Anyhow, Assert.jsm supports pluggable reporters, specifically meant to support any style. Everything will stay the way it currently is. > > -Boris > _______________________________________________ > dev-platform mailing list > dev-platform@lists.mozilla.org > https://lists.mozilla.org/listinfo/dev-platform _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
