Adopting the mozilla-central superreview policy in comm-central
Justin Wood (Callek)
callek at gmail.com
Sun Jun 13 02:05:51 UTC 2010
On 6/12/2010 5:52 AM, Mark Banner wrote:
> On 12/06/2010 03:00, Justin Wood (Callek) wrote:
>> On 6/11/2010 1:39 PM, Dan Mosedale wrote:
>>> On 6/10/10 4:59 PM, Justin Wood (Callek) wrote:
>>>> Now I know TB has required tests, but we are still in a way "O,
>>>> orange; must be a perma orange, likely not my fault" general
>>>> mentality on trunk.
>>> Now that 3.1 is in release candidate mode, the trunk is being
>>> whipped gradually back into shape.
>> Of course, but TB also runs much less tests than SeaMonkey, even with
>> its added MozMill tests (That SM does not yet run). I don't know of
>> any way to get TB to run the relevant tests reliably, but that does
>> make it less tests as well that are run relating to core code.
> I assume by "much less tests" you're referring to mochitest, reftest
> and all. If so, that has basically been a conscious decision that we
> do not need to run those tests constantly because Firefox is running
> them for us. Yes, some of the core code base is slightly different,
> and there may be one or two areas that running the tests against
> Thunderbird may reveal something, but on the whole, we're just letting
> the Firefox builders do the work for us there.
For 99.9% of the tests, I agree; for the other .1% I don't think the
effort to try and run them is worth it.
> There was talk about running reftests occasionally (because we can),
> e.g. near/at release times, I expect that may come once we do packaged
> tests, as it would be easier to do.
I _think_ reftests are the most likely to continue to pass between
Firefox and Thunderbird/SeaMonkey, due to reftests primarily testing
Gecko Rendering stuff not preferences/code of other parts of the app
> Re getting TB tests run reliably. AFAIK apart from one intermittent
> failure (bug 552804), all the MozMill tests for Thunderbird are pretty
> stable on Windows and Mac - Linux we know has various issues.
> xpcshell-tests should also be pretty stable. So if you're getting
> problems running tests, then that's something you should be filing a
> bug on (or a new thread) and getting it resolved.
I agree TB is in a MUCH MUCH better state than SM as far as reliable
tests, that said I am starting to find time to sneak in work on test
stability as well, which I am turning into bugs and patches where
appropriate. though I have not yet looked too deeply into MozMill tests
(which is the majority of TB ones).
~Justin Wood (Callek)
More information about the tb-planning