New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 842653 link

Starred by 4 users

Issue metadata

Status: WontFix
Owner:
Last visit > 30 days ago
Closed: Jun 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 1
Type: Bug-Regression
Proj-VR
Proj-XR
Proj-XR-VR



Sign in to add a comment

Navigating can stop UI elements from rendering.

Project Member Reported by vollick@chromium.org, May 14 2018

Issue description

I was able to repro by doing this
 1. Enter VR browsing
 2. Navigate forward and backward (by typing in the omnibox and clicking the back button).

Eventually, I got into a state where the content, the laser, the reticle and the floor grid weren't rendering. Some of the UI did render, though (eg, the controller).

Interestingly, when I hadn't downloaded the background component, when I got into this state I drew only a grey skybox (no gradient) and there were clearly visible seams between the faces.

This appears unrelated to the tree walk optimizations, but I haven't had a chance to bisect.
 
Screenshot_20180512-195301.png
228 KB View Download
Screenshot_20180512-195826.png
535 KB View Download
Owner: cjgrant@chromium.org
Status: Started (was: Available)
Picking this up.  I wonder if there's any chance it's the clipping change.  I'll try and repro with and without, for starters.
I haven't been able to reproduce this.  I'll keep trying while testing other work.
Owner: ----
Status: Available (was: Started)
Stopping work until someone can repro this again.
Owner: acondor@chromium.org
Status: Assigned (was: Available)
This issue was reported on the 14th, and clipping support landed May 11.  There was an issue with clip region hiding the platform UI texture that needed fixing.  

The laser is drawn using its own fragment shader, but is a BaseQuadRenderer and BaseRenderer under the hood, which were affected by the clipping CL.  It does't look like the laser could have been affected, but on closer inspection, BaseRenderer calls into the program to get the location of u_ClipRect, something not present in all elements (like the laser).  This is probably benign, but I wonder if any code should be DCHECKing on a valid clip rect uniform.

The timing relative to the clipping change just seems too suspect.  Aldo, can you see a direct link between the clipping issue and what Ian saw?




I cannot see the relationship. Querying a uniform that is not found in the shader is not a problem, even if you try to use the handle. Also, the bugfix was specific to external textures.
Owner: vollick@chromium.org
Ian, I assume there's no chance this was a local change, but I wasn't able to repro, and there's been no other sightings.  Was this readily reproducible for you?  I think we should close this off and reopen if it's seen again.

Assigning to you to disagree or close.
Status: WontFix (was: Assigned)
Yeah, if no one can repro, I think we should close it out, too.

Sign in to add a comment