Travis has been broken for a long time

[mgol]

Travis tests have been broken for over 4 months: https://travis-ci.org/jquery/sizzle/builds. That's bad, especially that new commits have been landing since and we can't know if they didn't introduce regressions.

[gibson042]

It's true, partly because of #362 and partly from the cross-window Sizzle.noConflict() testing (addressed by 087ba11) that I didn't notice through the noise of the former. I have been confirming fixes manually, and we all know how that goes.

Ideas and PRs welcome.

[mgol]

We reverted Karma to 0.12 so #362 shouldn't be an issue. We have timeouts in other browsers, updating karma-browserstack-launcher to the newest release might fix this problem; I wanted to do that & check but then I hit this noConflict issue. Perhaps we can re-try that now (unless the newest karma-browserstack-launcher requires Karma 0.13 to run but it doesn't seem to be the case).

[gibson042]

@mgol I've done what I can, and we're back to timeouts galore: https://travis-ci.org/jquery/sizzle/builds/108436309

Upgrading karma may help things, and v0.13.19 includes a socket.io-client that works with IE<8, but I guess we're still waiting for forceJSONP.

[dignifiedquire]

I just released 0.13.20 which ships with forceJSONP

[gibson042]

Fixed by 85123db: https://travis-ci.org/jquery/sizzle/builds/109198885

Thanks @dignifiedquire!