redo #2991: extend PR #2985 (moved canvas init) to also solve issue #2967 (vr2vr traversal in Oculus Browser) #2996
Conversation
|
rebase needed to remove the commit that has already landed |
machenmusik
force-pushed
the
canvasInit
branch
from
c0324f5
to
0c1b945
Compare
|
rebase done - are you in a position to sanity check? |
|
I tried quickly but it does not seem to work for me. I did not have the controller though (I hacked the link traversal demo with a gaze based cursor). Tomorrow I'll try with the proper controller again |
|
Works for me with this test page as example https://s.codepen.io/machenmusik/debug/XaVMWJ/mVAbGERyxabk Once you enter VR, it reloads in 5 seconds. So if you end up out of VR, you know it is not working. |
|
I'm trying the link traversal example with your branch and the controller and I get dropped of vr. Does |
|
for me, it is working for simpler scenes, but not for ones with big assets to download -- I suspect it runs into the timeout issue in those cases |
machenmusik
force-pushed
the
canvasInit
branch
from
7346c30
to
4df1a9a
Compare
|
looks like the setTimeout from prior PR is still needed for the VREffect to be available in the heavier cases. |
machenmusik
force-pushed
the
canvasInit
branch
from
4df1a9a
to
14ff5a6
Compare
|
@dmarcos can you confirm latest PR version works for you? |
| // This event may also fire in other circumstances (e.g. putting on the Rift headset). | ||
| window.addEventListener('vrdisplayactivate', function (evt) { | ||
| var s = window.AFRAME.scenes[0]; | ||
| if (s.effect) { s.enterVR(); } else { setTimeout(s.enterVRBound); } |
There was a problem hiding this comment.
I don't think it is guaranteed that in the next tick the effect is going to be defined. I'm trying to find a more robust way to do this.
There was a problem hiding this comment.
Also is not what I would expect the event to behave following the spec. The user action policy should not apply to the next tick logic.
There was a problem hiding this comment.
If you find a more robust way, great. I tried several things and did not find another way for gesture permission to survive. See WICG/interventions#12
There was a problem hiding this comment.
(More specifically, WICG/interventions#12 (comment))
|
@dmarcos any progress on a more robust solution? (Where this.effect will definitely exist while preserving user gesture?) |
|
@machenmusik out of curiosity, which example(s) are you testing with this? just https://aframe.io/aframe/examples/showcase/link-traversal/, yeah? |
|
This will be handled in un upcoming PR implementing a loading screen |
|
@cvan perhaps irrelevant now, but link-traversal example, this codepen (https://s.codepen.io/machenmusik/debug/XaVMWJ/mVAbGERyxabk) and a non-public app |
|
@dmarcos ok, as we discussed at OC4 |
|
@machenmusik thanks for the conversation at OC4 |
|
@machenmusik that CodePen URL gave me a 403. just posting this in case anyone else's reading: https://codepen.io/machenmusik/pen/XaVMWJ/ |
|
Thanks and sorry @cvan! |
|
no sweat! I was just testing WebVR flows in Oculus Browser. and I know you were working on this, so wanted to check. thanks for working on this. before I do a deeper dive, are there any known issues you know of for A-Frame + Gear VR? I ask because the new Oculus headsets are coming soon, and I want to make sure A-Frame and three.js are prepared. thanks again for all your PRs and keeping your finger on the pulse 👍 |
per @dmarcos:
I merged by mistake and reverted. Sorry for that. Can you please reopen? #2985 has landed so we need to rebase
(Rebase forthcoming.)