View directives on error

[jwasilgeo]

@tomwayson please review.

I actually had to do the following in Chrome (48.0.2564.109):
chrome://flags/: Disable 3D software rasterizer (yes, do this)
chrome://settings/: Use hardware acceleration when available (you might also have to un-check this)

Resolves #239
Resolves #240

I tried to split different concerns into different commits:

• I added new info about the on-error to the patterns page about the other events, but perhaps that wasn't the perfect place to dump in that new info.
• Also, I don't know of a good way to functionally e2e test the addition of the on-error event to the view directives test pages, so only the site examples currently benefit from this.

[jwasilgeo]

PS: In an attempt to see these pending changes on my mobile devide, I was not entirely sure if JSAPI was rejecting my scene view promises as suggested by this scene webgl support example. It may be the case that something else is being used in JSAPI docs when navigating to examples on unsupported devices.

[tomwayson]

If I go to that link on my phone (Chrome on Android), or go directly to the link from the view live sample button on that page, I see the appropriate error messaage (WebGL is not supported on your platform/browser). I also see Chrome's own "Rats! WebGL hit a snag" message. Particularly in the case of the live sample, I don't think this can come from any other code other than than what's in the sample itself. So, ultimately we should be able to reproduce that behaviour in our site.

I'll give this a shot later tonight. In the mean time, if you are able to deploy the updated site to your own fork's gh-pages branch, send the link and I'll test from my phone.

[tomwayson]

@jwasil

excellent work on a tough problem. I tested using chrome remote debugging on my phone and also pushed to my fork's gh-pages, and it seems to me that sometimes (usually the first time?) I load any example page w/ a web scene on my phone, I don't see the error message that I should, but subsequent times I do.

At first I thought maybe we do need $scope.$apply() in the promise's errback, so I tried that and saw all the already in digest nonsense, so I undid that.

I tried a few other things, including setting the view's div to display:none in the errback (as the JSAPI sample does), but I could not reproduce the problem (of not seeing the error message).

Anyway, I think what you have here is the best we can do for now, so I've merged and deployed gh-pages. We'll just have to keep an eye on this and see if anything comes up.

[jwasilgeo]

Thanks for the very thorough review and help with investigating this PR. Hopefully this will take care of it, but yes, we may need to revisit one day. Thanks!