mobileFileBrowser demo is broken on trunk

[Ming Zhe Huang]

Please check: ​http://archive.dojotoolkit.org/nightly/checkout/demos/mobileFileBrowser/demo.html

Click the list item and you'll see error in browser debug console. Probably it's caused by the refactoring of some fundamental dijit API.

[bill]

Good catch, that demo is missing a dependency, I'll fix.

[bill]

In [26207]:

add in missing dependencies, test still doesn't work though, refs #13688

[bill]

Oof, the problem is deeper than I thought. Rawld, can you take a look?

"has" somehow gets defined as {connect: ..., disconnect: ...} rather than what the has.js factory returns, and when I start putting in debug statements then kernel and or lang are defined as "not-a-module" rather than hashes. Apparently kernel.js and lang.js are getting loaded but the factory functions aren't executing.

This is from testing on FF5/mac.

[bill]

Note that the demo (actually, all the demos) have two <script> tags, one for dojo.js and one for src.js. src.js isn't an AMD module, but rather plain javascript. But still might be connected, since dojo.js has presumably already internally started firing off require()'s for lang.js and maybe parser.js.

[Rawld Gill]

Likely fixed as of [26215]; checked src and built version and it seems OK. Please recheck and verify that this is fixed.

[bill]

I'm still getting intermittent errors against the source build about has.add() not being defined. Tested on FF5/mac running against apache.

I can see this is a race condition so I can turn on tracing if you need me to, and you remind me how.

[Chris Mitchell]

In [26266]:

refs #13688 converted dojo/fx easing and Toggler modules. Updated dojo/fx.js to automatically define Toggler and baseFx apis for compat.

[Ming Zhe Huang]

I just tested with the latest trunk codes. It works.

[bill]

I'm still seeing the error on the latest code. I needed to run it four times though before the error popped up. Like I said it's a race condition.

[Chris Mitchell]

disregard my patch above re: Toggler and easing...i typo'd the ticket #...should have been #13699.

[Douglas Hays]

This fails every time for me on Chrome(13)/win, from both a demos-all build or against source.

[Chris Mitchell]

this hadn't been fully converted to amd with the latest dojo api changes. i just checked in a fix r26361. We should prob be able to close this now.

[bill]

The worrisome thing is why it was failing before. Looks like a bug in the loader?

If it was just a question of missing dependencies then that's great as it doesn't indicate a loader timing bug; but I don't see how the missing dependencies would cause has to point to something else etc.

[Douglas Hays]

working for me

[Rawld Gill]

Replying to bill:

Note that the demo (actually, all the demos) have two <script> tags, one for dojo.js and one for src.js. src.js isn't an AMD module, but rather plain javascript. But still might be connected, since dojo.js has presumably already internally started firing off require()'s for lang.js and maybe parser.js.

Any projects that inject code via explicit <script> tags can expect to work in async mode if and only if the injected source uses AMD require. As of [26381], that is the case with mobileFileBrowser.

[Rawld Gill]

Replying to bill:

The worrisome thing is why it was failing before. Looks like a bug in the loader?

If it was just a question of missing dependencies then that's great as it doesn't indicate a loader timing bug; but I don't see how the missing dependencies would cause has to point to something else etc.

I agree it's worrisome. Unfortunately I'm unable to repo to bug in my environment (I tried against [26265] on FF 5/linux; chrome 13/in). So I'm going to close it for now.

[bill]

I figured out the bug, filed it as #13768. You are off the hook :-), the dojox/mobile code is doing something it shouldn't.