1248525 - JS Flot bug with Firefox 44.0.2

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[moffatjd]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:44.0) Gecko/20100101 Firefox/44.0
Build ID: 20160210153822

Steps to reproduce:

When I updated to 44.0.2, my javascript plotting using JS Flot stopped working as expected.

I reverted to Firefox 43.0.4 and the problem is not there.

I updated back to 44.0.2 and the problem returns.


Actual results:

See the following jsfiddle:

http://jsfiddle.net/moffatjd/4c6f6h4w/2/

There should be no shading.


Expected results:

Both Safari, Chrome, and Firefox 43.0.4 all behave as expected (no shading). But 44.0.2 seems to have broken the functionality and the shading appears. Note, I that I don't think it was broken in 44.0.1. It was very recent. 

I know this may seem like a small bug, but it's a big deal to me, and I'm guessing that it might be indicative of some other underlying javascript functionality. Perhaps this will shed some light.

Comment 1

[Tooru Fujisawa [:arai]]

it's already fixed in firefox 46, but firefox 44 and 45 are affected.

Regression range (probably bug 1150944):
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=e5746165c704a82fc92d059ef9303562d6b3c886&tochange=5ed72fdd632767e03d6862efd10fc15a34256845

and fixed by bug 1082598:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=d8e260e1535f0ff3877b21d6846380dedf449453&tochange=6630a176477ef903dcf3057db755abc43ba5018d

Comment 2

[Tooru Fujisawa [:arai]]

Attachment: Reduced testcase for wrong filling

Here's a reduced testcase.

ctx.fillStyle = "red";
ctx.beginPath();
ctx.moveTo(0, 0);
ctx.lineTo(0, 30);
ctx.lineTo(30, 30);
ctx.lineTo(30, 30);
ctx.lineTo(0, 30);
ctx.fill();

In Firefox 44.0.2 on OSX, it fills tetragon (0, 0), (0, 30), (30, 30), (10, 20).
This happens only when gfx.canvas.azure.accelerated=true.

mattwoodrow, can this be fixed somehow?
(turning off Skia or backport bug 1082598 or something...?)

Comment 3

[Matt Woodrow (:mattwoodrow)]

We only have a few weeks left until Firefox 45 ships, I'm not sure we can do much at this point unfortunately.

Uplifting a skia update is terrifying, the only option would be to disable skia-gl.

Comment 4

[u279076]

I agree with Matt's risk assessment in comment 3. I might be more inclined to accept it if this was a stability or security problem but it's not. In the meantime anyone experiencing this issue (assuming it's mostly developers) can switch to Developer Edition.

Comment 5

[moffatjd]

Thanks all. I know little about your risk assessment, but I empathize. This specific problem does impact users (and not just developers), but we can come up with a temporary workaround until 46 ships. Thanks again.

Comment 6

[Sylvestre Ledru [:Sylvestre]]

We are going to let it ride the train from 46 then...

Comment 7

[Lee Salzman [:lsalzman]]

(In reply to Sylvestre Ledru [:sylvestre] from comment #6)
> We are going to let it ride the train from 46 then...

I did do some investigation of this, but the amount of changes to the way Skia handles paths internally were too significant to isolate any easy fix. So I am in agreement in just letting this one slip until 46 comes along.