Author: intrigeri Date: To: The Tails public development discussion list Subject: Re: [Tails-dev] [review'n'merge: 1.3]
bugfix/7951-refactor-chroot-browsers
anonym wrote (30 Nov 2014 20:42:04 GMT) : > On 31/10/14 16:02, intrigeri wrote:
>> Ooops, I forgot this one. Obviously the test fails, so I think I've
>> fixed it in e44f97c.
>>
>> No big deal, but it makes me question how much trust I can put, as
>> a reviewer, into "I've tested it (both manually and automatically)
>> very thoroughly" => how much change did happen on this branch *after*
>> you tested it thoroughly? > I have no good explanation for this. The test suite ran well for me but
> I may have focused more on the Unsafe Browser.
Fair enough.
> Any way, let's interpret that quote lightly, and in any case the
> merger should test the stuff to their satisfaction on their own,
> right? :)
In theory, yes. In practice, given how resources-consuming the
automated test suite is, and the fact we currently have no way to
externalize its running, frankly I'm not keen to run again features
that the branch developer tells me they've run successfully: here,
I basically can't build an ISO while running the test suite without
harming the reliability of the latter, so it's partly blocking me from
working on anything else. (TODO++ upgrade that Mac thing lying here
and set it up to run the test suite :)
What I try to do when reviewing is to run all the *other* features, to
catch unintended breakage at a distance. So, for the time being,
indeed I tend to rely on "I've run $FEATURES successfully". Can we
agree that this implicitly means "on the exact commit that I'm
proposing to be merged" in the future?
(And hopefully, in 2015 we'll get the infra we need to avoid having to
run the test suite by hand, and all this discussion will be moot :)
