Signal execution seems overly delicate

[pjt33]

If I tweak examples/Avatar/Avatar1.json to remove all the parts except Sword1 then the demo still works. If I then remove Sword1 as well, leaving "parts": [], the demo fails. By inserting breakpoints, I see that with Sword1 the meshLoaded signal fires and the animations are loaded before the avatarLoaded signal, but without Sword1 the meshLoaded signal fires and the animations are loaded after the avatarLoaded signal.

It could be that this is intended, but the docs do say

Once the Avatar component has finished loading the description, has instantiated the Mesh component(s) and their mesh assets have loaded, the character is ready to use for animation playback. At this point the Avatar component will emit a signal, avatarLoaded, to which application code can hook up.

so it seems to me that it's a bug.

[antont]

Sounds like a bug, yes, thanks for reporting.

I think is good to check how the new / different code in the dev2 (planned to be 2.0 for next release soon) works in this respect -- the avatar component is just a stub there now though in https://github.com/realXtend/WebTundra/blob/dev2/src/entity-components/EC_Avatar.js

If you don't need the parts & other avatar configuration business, is fine to use a normal mesh if you just need one with skeletal animation.

[jonnenauha]

Proper implementation of EC_Avatar.js is coming. Also related parts for .xml/.avatar Assets and parsing them are coming too. This is a lot of code as it brings in avatar attachement and related changes to the skeleton asset business.

@cvetan5 labeling this as todo for 2.x

Edit: Looks like i cant create new labels here!? :)