On Sat, Mar 31 2012, Mark Walters <markwalters1009@gmail.com> wrote: > It seems off to call is success without checking that the value has > actually been set. Of course it is checked in the notmuch config list > test introduced in the next commit but I think if it would be better to > check with notmuch config get here too (i.e. check that reading back the > value gives what you want). Otherwise a failure in `setting' will show up > as a test failure in `listing'. Hey, Peter. I think Mark makes a good point here. I think it would make more sense for the test to set the value, and then check that the value is properly set as expected. It would make the tests multi-step, but that's fine. There's plenty of precedent for that. jamie.