1479947 - CanvasRenderingContext2D.drawImage() doesn't work with video element as image source

Status: RESOLVED DUPLICATE of bug 1486659

(Core :: Graphics: Canvas2D, defect, P3)

Description

[alistair]

User Agent: Mozilla/5.0 (Android 8.1.0; Mobile; rv:61.0) Gecko/61.0 Firefox/61.0
Build ID: 20180621125625

Steps to reproduce:

Attempt to copy video frames to canvas context using CanvasRenderingContext2D.drawImage() as in these examples http://html5doctor.com/video-canvas-magic/


Actual results:

Canvas stays blank


Expected results:

Video frames copied to canvas

Comment 2

[Kevin Brosnan [Ex-Mozilla]]

Not a dupe.

Comment 3

[Kevin Brosnan [Ex-Mozilla]]

This seems to be broken on every device I've tried.

Comment 4

[Andrew Osmond [:aosmond] (he/him)]

Works fine on desktop, but I guess it is limited to Android. jnicol, are you able to reproduce with one of the demos, e.g.:

http://html5doctor.com/demos/video-canvas-magic/demo2.html

Comment 5

[Jamie Nicol [:jnicol]]

Yeah, this is a known issue. The problem is that on android we decode the video frames in to a surfacetexture, which only allows one gl context to consume it at a time. But we need 2, one for reading the pixels for the canvas, and another for compositing.

In e10s this is very hard (see Bug 1486659), because they are in different processes. For non-e10s this could be solvable, with perhaps a lock of some kind. But I don't know if we want to solve it differently for non-e10s than for e10s. Snorp?

Comment 6

[James Willcox (:snorp) ([e-mail]) (he/him)]

Yeah, I think I'm going to dup this to Bug 1486659 since it's essentially the same problem. I tried to use a lock allow us to detach the ST from the Compositor so we could solve this bug, but it didn't really work. We had corrupted images on some devices.

Comment 7

[alistair]

Bug 1486659 has been closed fixed in firefox66 upwards, but the drawImage issue remains in firefox68 on my Pixel 2