On Mon, Oct 08, 2012 at 12:05:58PM -0700, Gregory Szorc wrote: > We now have a tool in mozilla-central that has a much better UX for > running tests (mach). It's not perfect yet, but it's getting there > (please write patches!). > > The build peers (or at least a few of us) really don't like the make > targets to run tests because they are awkward, both to maintain and > for people to use. mach and having Python drive everything offer > compelling advantages over simple make targets.
I like the make targets. > The make targets for running tests aren't used by buildbot (well, at > least a lot of them - there might be a straggler or two). That the make targets are not used to run tests on buildbots doesn't say much about the usefulness of these targets. As a matter of fact, I've always thought buildbots should be using these targets. > Putting this all together, the stage is set to remove the make > targets for running tests. > > I'm writing this post to see what obstacles/resistance there are to > removing the make targets for running tests. Obviously a > prerequisite is having mach reach feature parity with the make > targets. What other concerns are there? Besides the points raised by other people in this thread, I'll mention that running tests in a debugger is currently not supported by mach, and that's a clear show-stopper (I'm pretty sure there's an open bug for that). That being said, I do appreciate the mach commands to run tests, they are simpler than the make targets to run individual tests. They have weird "error" handling, though, but that comes from the test harness (for instance, running mach xpcshell-test toolkit/devtools/debugger/tests/unit/non-existing-test.js runs all the tests in toolkit/devtools/debugger/tests/unit, instead of failing) Cheers, Mike _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
