Open Bug 1333495 Opened 7 years ago Updated 2 years ago

Fix and reenable various webgl tests that use random images on the web

Categories

(Core :: Graphics: CanvasWebGL, task, P3)

task

Tracking

()

REOPENED
Tracking Status
firefox54 --- affected

People

(Reporter: bzbarsky, Unassigned)

Details

(Whiteboard: [gfx-noted])

The webgl testsuite uses random images on the web.  This is totally busted, and I filed https://github.com/KhronosGroup/WebGL/issues/2273 on that, but in the meantime...

What we should probably do is add the two hostnames involved to our test HTTP server and map the relevant paths to something in-tree.  Then I expect we can remove several "skip-if = 1" from dom/canvas/test/webgl-conf/generated-mochitest.ini as well as the unconditional CRASH annotations from <https://dxr.mozilla.org/mozilla-central/search?q=path%3Atesting%2Fweb-platform%2Fmeta+CRASH&redirect=false> -- those crashes are because access to random web servers during tests kills the process.

Doing this would improve our test coverage, which is a good thing, right?  ;)
Flags: needinfo?(jgilbert)
Priority: -- → P3
Whiteboard: [gfx-noted]
These images are now hosted on get.webgl.org, so it's a matter of mapping that domain in our test infra and making the relevant path point to the right images...

This was fixed.

Status: NEW → RESOLVED
Closed: 5 years ago
Flags: needinfo?(jgilbert)
Resolution: --- → FIXED

How is this fixed? As far as I can tell, still have tests in our tree that are truing to load stuff from https://get.webgl.org, which kills the test, so we are skipping them instead of running them.

Status: RESOLVED → REOPENED
Flags: needinfo?(jgilbert)
Resolution: FIXED → ---

I thought we fixed this in a PR to the WebGL repo.

I'm not particularly worried about this, and I don't know the path forward to fixing this, but we can leave this open.

Severity: normal → minor
Type: defect → task
Flags: needinfo?(jgilbert)

I thought we fixed this in a PR to the WebGL repo.

That fixed the tests pulling in a random image from somewhere on the web that wasn't even under the control of Khronos.

I'm not particularly worried about this

OK, if we're ok with the missing test coverage.

and I don't know the path forward to fixing this

Comment 0 generally describes the path, no?

Flags: needinfo?(jgilbert)

What we should probably do is add the two hostnames involved to our test HTTP server and map the relevant paths to something in-tree.

If you want to volunteer, sure, but this all sounds like whole days of work trying to track someone down to figure out how to do these things! I don't know how to move forward on those things.

Flags: needinfo?(jgilbert)
Severity: minor → S4
You need to log in before you can comment on or make changes to this bug.