1271090 - Uncorrect SVG feFuncA behavior

Status: RESOLVED WORKSFORME

(Core :: SVG, defect)

Description

[dom1n1k]

Attachment: Left side - buggy result by FF, right side - normal result by Chrome

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/50.0.2661.94 Safari/537.36

Steps to reproduce:

There is gradient, where alpha change from 0 to 1.
Apply filter feComponentTransfer, where
<feFuncA type="linear" slope="10" intercept="0"></feFuncA>



Actual results:

Something happens with RGB-components. Visible posterization. See attached screenshots and test example:
http://codepen.io/dom1n1k/pen/qZgrpr?editors=1100



Expected results:

Slope 10 means to obtain values greater than 1 - they should be clamped (obviously).
Other channels (RGB) should not be touched.

Chrome and IE works fine.

Spec: https://www.w3.org/TR/SVG/filters.html#feComponentTransferElement

Comment 2

[dom1n1k]

I'm not sure that this is a duplicate.

Main question not about banding. Main question about influence feFuncA to other channels.
I guess overflow after multiplication alpha channel to large multiplier (slope).

Sorry for my bad english :)

Comment 3

[Robert Longson [:longsonr]]

I see banding on the left hand side. I'm not really sure I understand what else you're trying to say here.

Comment 4

[Robert Longson [:longsonr]]

Perhaps you could try to create a more obvious testcase?

Comment 5

[dom1n1k]

Simplified example: http://codepen.io/dom1n1k/pen/ONdjWE

1. In gradient color is constant (orange). Alpha from 0 to 1.
2. Filter should change alpha channel only.
3. But we see changes in other channels (yellow instead orange).

I suspect overflow. Banding - side effect.

Comment 6

[Robert Longson [:longsonr]]

Firefox does not support the color-interpolation CSS property. If I remove that Chrome and Firefox look the same. Is that the issue, if so we already have a color-interpolation bug.

Comment 7

[dom1n1k]

Hmm.
If I delete attributes color-interpolation and color-interpolation-filters, result is slightly different, but still obviously wrong. And not the same with Chrome.
feFuncA should not change color (only alpha). But color was changed.

Comment 8

[Robert Longson [:longsonr]]

obviously wrong how?

Comment 9

[dom1n1k]

I attached new picture.
Chrome and IE look slightly different, but within rounding errors.
Firefox - clearly different color.
(testcase - http://codepen.io/dom1n1k/pen/ONdjWE)

Comment 10

[dom1n1k]

Attachment: See comment #9

Comment 11

[Robert Longson [:longsonr]]

Firefox seems to display the same as Chrome now.