1364140 - canvas drawImage ignores clipping for various globalCompositeOperation values (regression)

Status: UNCONFIRMED

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

Description

[Daniel Szabo]

Attachment: globalCompositeOperationVSclip.html

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:54.0) Gecko/20100101 Firefox/54.0
Build ID: 20170508140042

Steps to reproduce:

Open the attached html file. It will display images of canvas drawImage operations using different globalCompositeOperation settings.
For each globalCompositeOperation, the top image is the result of the operation without clipping applied on an alpha-enabled canvas.
The image in the middle is the result of the operation using a smaller clipping rectangle.
The bottom image is the result of the operation using clipping, however, applied on an alpha-disabled canvas.


Actual results:

Clipping is ignored in the following cases:
- "difference" regardless of alpha setting
- "source-atop", "destination-out" and "lighter" used on an alpha-disabled canvas

Moreover, transparent areas will be painted black instead of dark blue on alpha-disabled canvases.


Expected results:

Clipping should work in all cases.
Transparent areas should be painted dark-blue on alpha-disabled canvases.

This is a regression:
- FF28 renders all cases correctly.
- FF37 messes up other cases totally, however, clipping is working, when an image is rendered.
- FF44 renders clipped images correctly, however, transparent areas are painted black on alpha-disabled canvases.
- FF52 (and FF54) renders several cases without clipping.

Comment 1

[Daniel Szabo]

Attachment: Rendering results of FF28

Comment 2

[Daniel Szabo]

Attachment: Rendering results of FF37

Comment 3

[Daniel Szabo]

Attachment: Rendering results of FF44

Comment 4

[Daniel Szabo]

Attachment: Rendering results of FF52

Comment 5

[Loic]

Regression range for the clipping issue:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=494289c72ba3997183e7b5beaca3e0447ecaf96d&tochange=538d248fa252a4100082fd9bc3fdc08d322cda22

Comment 6

[Jeff Muizelaar [:jrmuizel]]

It would be nice if to have a narrower regression range.

Comment 7

[Daniel Szabo]

Actually, the clipping issue started in FF46

Comment 8

[Jeff Muizelaar [:jrmuizel]]

It looks like with D2D we get different results than Edge and than with Skia.

Comment 9

[Bas Schouten (:bas.schouten)]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #8)
> It looks like with D2D we get different results than Edge and than with Skia.

I believe the regression was from the bug where we fixed some drawing issue with complex blend modes. It's noted in one of the other filings of this bug.

Comment 10

[Jeff Muizelaar [:jrmuizel]]

Can you link to it from here?

Comment 11

[Bas Schouten (:bas.schouten)]

I think this is bug 1333090.

Comment 12

[Beliov]

+1