Last modified: 2014-08-22 10:31:44 UTC
Urls like, https://en.wikipedia.org/wiki/Mike_Godwin#mediaviewer/File:T<ess.pdf Result in a completely black page with discoverable way to get out and no commmunication as to what happened. The console shows: Uncaught Error: Unable to parse title VM299:35 Title VM299:35 RP.parseHash RP.parseLocation VM320:90 MMVP.hash VM320:103 deferred MMVB.isCSSReady Presumably there is an instance of `new mw.Title` where it should use mw.Title.newFromText with a condition check to verify the title was not invalid. It is user input after all, and even if it would try to match the DOM element first, the DOM can also contain arbitrary elements that are made to look like mediawiki image frames, and in addition, title validation rules can change over time, and in addition, the js class is only an approximation to the sever-side variant. There's always a risk of it being looser or stricter, so always validate titles (unless maybe they come from the API or an mw.config value directly). The page being completely black is more common actually, since it's the standard thing that happens if there's an uncaught exception anywhere. Looks like maybe the UI and logic aren't separated properly that it causes the UI to get stuck half-way when it's doing what looks like simple logic. It should probably set up more of the UI at once (e.g. don't put up a black dimmer without also, in the same atomic dom change, provide a way to get out of it).
