On 06/12/2012 15:01, Sébastien Santoro wrote:
On Thu, Dec 6, 2012 at 10:52 PM, Tim Landscheidt tim@tim-landscheidt.de wrote:
That is all. The slow unit tests are no more run on patchset submission.
We really need them.
The tests philosophy is "there is an issue with this change".
The Jenkins philosophy is "automate as most as possible to discover them as soon as possible".
Furthermore, one of the current Wikimedia politics is "get more technical contributors" and do outreach tout azimut. The tests are really important there, as they don't know each particularities of each class of the code (and I'm including myself in this area, there are MediaWiki classes I haven't get the opportunity to read yet).
Should I really put logical constructors between these facts or the "we really need tests to run" is evident for all?
It should be - as a volunteer it can be quite difficult to find someone to review a change, and that is after it already passes the basic tests. Asking folks to review things when we don't even know if it passes the basics is not going to make the process go any easier or faster, and having to wait even longer and poke even harder is the sort of thing that drives people to give up on the entire process.