Great! If they pass - why not put them on master?
On Tue, Nov 12, 2013 at 5:52 PM, Braden Shepherdson <bra...@chromium.org>wrote: > Following up here with the note that there's a public branch called > e2e-tests with working tests for create and platform commands. These tests > are touching the real filesystem using only slightly hacked-up fixtures. > > A couple of things we learned: the res/ directory in the helloworld > template is 3MB, the rest of the project including Android source and > everything is about 280K. So drop the res/ directory from the fixtures, now > it's much faster. > > We're having to inject into our freshly created project a config.json > pointing at the fixtures instead of letting it do the real lazy-load, but > that's working pretty well. They're fast, too, faster than some of the > "unit" tests. Hopefully that trend will continue. > > Braden > > > On Tue, Nov 12, 2013 at 4:56 PM, Anis KADRI <anis.ka...@gmail.com> wrote: > > > Wow... > > > > I meant > > > > "replace them with real world assertions (that touch the FS)". :-) > > > > On Tue, Nov 12, 2013 at 1:55 PM, Anis KADRI <anis.ka...@gmail.com> > wrote: > > > On Tue, Nov 12, 2013 at 1:49 PM, Andrew Grieve <agri...@chromium.org> > > wrote: > > >> Want to give a big +1 to deleting tests that are testing > implementation > > >> rather than results. Having such tests is a real drain on > productivity. > > > > > > I tend to agree with this statement. Maybe not delete but add/replace > > > them real world assertions to them (that touch the FS). In some cases > > > it makes sense to test the implementation to make sure that an action > > > goes through the proper flow. > > >
