Integration tests and CKEditor

[vmassol]

ok didn’t notice there were two topics but I don’t think it changes anything: either the feature (comments, mentions, etc) is doing something special with CKEditor in the way it interacts with it (it has some JS code, etc) and then it makes sense to test with CK to exercise that code, either it doesn’t and this case I don’t think it’s necessary or good to test with CK (extra test time, no strategy/consistency with other tests, etc).

@surli

To give you a very simple example: CKEditor when loaded is performing some checks by loading an editor and destroying it afterwards. Those checks are triggering some JS focus events on the page. Those focus events were directly responsible for some bugs in annotation reply, because the page was not properly scrolled when replying to an annotation.
You cannot find those kind of interactions without testing really the integration.

These focus events are inside the comments or annotation module?

In any case, I agree that you need to test all the infinite combination of items to be really sure. But you cannot. You’re mentioning this after the fact (i.e. you know the problem). But what if there’s an issue with the wiki editor now? or when annotation/comments is combined with mentions (and that’s not tested since we test feature by feature in isolation), etc? So it’s about probability and what we can test in timespan we have (ie with the number of commits we do every day and the number of agents we have).