CloudFlare Rocket Loader must DIAF

[classilla]

Uplifted from https://tenfourfox.tenderapp.com/discussions/problems/8670-site-login-failure

After debugging, the problem with things not loading on the page appears to be because Rocket Loader doesn't load them. I can't find an uncompressed version and testing the site in the Debugger doesn't indicate anything obvious.

The workaround is to relax what is accepted for <script type="" so that the browser will just load the scripts for us and Rocket Loader can die in a fire.

[classilla]

I don't like the GetAttribute hack -- I'd rather deal with this at the parser level -- but parsing is so asynchronous this seems the most straightforward way to deal with this issue. It does work properly on the indicated site. The hack is needed to prevent Rocket Loader from loading scripts twice, which seems to happen sometimes, implying the actual cause is a race condition.

[classilla]

Closing, as other Rocket Loader sites are now confirmed working, and the original reported site seems to work fine here.

[chris-chtrusch]

I'm still unable to comment on the development blog, so I'll attach this here: Might TFF not being able to load images at https://vsco.co/feed have something to do with the cloudflare problem? Or is this a new issue? Checked with TFF FPR9 final, fresh profile, different user agents, JS/ion/baseline disabled etc. Console gives lots of CSS errors which I can't interpret as significant or insignificant. All I get is a spinning wheel where Safari 5/LWK starts loading images right away.

[classilla]

No, this is something else. It looks like a JavaScript feature we don't currently support, but it's minified, so it's very hard for me to figure out exactly what it wants.

[classilla]

I figured it out. It's async and await which we don't support and would need major surgery to do so. #490 was a thought about how to fake it out but I'm not so sure how workable it is on reconsideration.

[classilla]

Reopening; https://www.tampabay.com/mugshots/ still doesn't work with this hack, but does work with an IE8/11 UA. An exhaustive diff shows that the only difference in what is served is the Rocket Loader portions; the scripts and everything else are identical, so it must be we have an edge case. It may be better just to suppress Rocket Loader from loading altogether rather than suppressing the property query when it does, because in the IE8/11 case, the browser doesn't get served it.

[classilla]

This does seem to fix the problem.