Limit all test suite selection operations to #qunit-fixture

[gibson042]

For those that are explicitly document-scope, wrap with supportjQuery( $jQueryResults ).filter( "#qunit-fixture *" ) (cf. #2877 (comment) ).

[markelog]

I believe i have fixed that, there might be more, but they are not causing issues at the moment.

See b97c8d3 f1300f1

Note: that should be done for sizzle too.

[gibson042]

Why did you close this? It is a general hygiene ticket that should stay open until we are reasonably confident that there are no "others".

[markelog]

I believe i have addressed your comment mentioned in this ticket description, aside from resolved cases there is no proof that such tests even exist, they might, but it is not determined, so i think we reached "reasonable" criteria, otherwise i don't understand how you want determine if those selectors are there.

Also, i think bad hygiene is also to have tickets that are open forever, without any possibility of closing except for subjective feeling of confidence.

However, if you strongly feel about this, please feel free to re-open.

[gibson042]

i don't understand how you want determine if those selectors are there

By looking through all the unit tests.

i think bad hygiene is also to have tickets that are open forever, without any possibility of closing except for subjective feeling of confidence.

I agree, but closing it now is indistinguishable from "wontfix". The point here is to avoid re-occurrence, and if you're looking for an objective characterization, here you go:

• Someone comments on this issue certifying that they have examined all unit tests and found no selections scoped so broadly that they could match QUnit output, or
• A commit lands on the master branch claiming to correct all such selections

[gibson042]

• Someone comments on this issue certifying that they have examined all unit tests and found no selections scoped so broadly that they could match QUnit output, or
• A commit lands on the master branch claiming to correct all such selections

The first of these should be discounted, because I've already found some offending examples:

• attributes.js#L223
• attributes.js#L261
• attributes.js#L494-L496
• core.js#L769
• core.js#L777
• css.js#L482
• css.js#L570

I'll stop there, but you get the point.

[markelog]

Even though for most of those tests it doesn't matter what element is taken, it should work for any of them, but now i understand your train of thought.

A commit lands on the master branch claiming to correct all such selections

I would propose to do it gradually, since it could be a lot of work to do in action, one commit would be preferable, yeah, but it hard to do, for example i can take attributes module tonight.

I agree, but closing it now is indistinguishable from "wontfix". The point here is to avoid re-occurrence, and if you're looking for an objective characterization

Well, i always comment about such tests and fix a lot of them already, not even counting those aforementioned commits, so i think we no longer write such fragile tests.