619834 - Gradient with colors with alpha<1 are not smooth

Status: RESOLVED DUPLICATE of bug 627771

(Core :: Graphics, defect)

Description

[gadjo]

User-Agent:       Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b9pre) Gecko/20101216 Firefox/4.0b9pre
Build Identifier: 

Gradient from rgba(0,0,0,0.8) to #fff should look the same as from rgb(51,51,51) to #fff, but it doesn't (it has visible circles), see the testcase

Reproducible: Always

Comment 1

[gadjo]

Attachment: testcase

Click the mouse to see different versions of radial gradient

Comment 2

[Bas Schouten (:bas.schouten)]

I'm also seeing this in 3.6 on Windows (although the effect is more strongly present there in the rgb(51,51,51) to #fff case, could you confirm this?

Comment 3

[gadjo]

On Fx3.6/WinXP the effect with rgba(0,0,0,0.8) seems to be slightly less visible, and with rgb(51, 51, 51) more visible than on Fx4.0/Win7. Both cases are more similar to each other in 3.6/XP, on 4.0/W7 there is bigger difference between them.

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Bug 591600 might help here, though it also may be hard to do.

Comment 5

[gadjo]

I'm not sure, this bugs refers to the quality of gradient (local color changes) and bug 591600 is more about computing stop-colors value as I understood.

I've checked the colors of the gradient in my testcase (rgba(0,0,0,0.8) to #fff) and starting from the center they are:

51, 52, 53, 52(!), 53, 54, 55, 56, 55(!), 56, ...

I guess it's more about the precision of calculations or something like this.

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

The origin of that imprecision *may* be that we're interpolating both alpha and color (yielding results in 0-255) and then multiplying the two together, rather than multiplying and then interpolating as the fix for that bug would do.

Comment 7

[gadjo]

Requesting blocking2.0 as background for tab modal prompts looks bad

Comment 8

[Joe Drew (not getting mail)]

Alexes, is this something you're worried about? Right now I don't think this blocks.

Comment 9

[Alex Limi (:limi) — Firefox UX Team]

The way to fix it if we can't get this fixed in time for the tab-modal prompts would be to switch to the other syntax, right?

If this isn't too complicated to fix, I'd of course appreciate if it got fixed in the proper location too. Don't think I would hold the release for it, though. :)

Comment 10

[Justin Dolske [:Dolske]]

If we can tweak the syntax to fix this for tab-modal prompts, that's fine.

File a new bug for that, please? And attaching a patch would be fantastic. (hint, hint! ;-)

See http://mxr.mozilla.org/mozilla-central/find?string=tabprompts.css

Comment 11

[gadjo]

Filed bug 623481 (sorry, no patch as I'm not sure what other syntax you are referring to)

Comment 12

[Alex Limi (:limi) — Firefox UX Team]

(In reply to comment #11)
> Filed bug 623481 (sorry, no patch as I'm not sure what other syntax you are
> referring to)

I might have misunderstood what the initial bug report was stating, I thought that one of the ways to express this had a bug, and the other one didn't. If that's not the case, the fix is not as straightforward.

Comment 13

[Carlo Alberto Ferraris]

Attachment: another testcase

Comment 14

[Carlo Alberto Ferraris]

Attachment: another testcase (screenshot)

Screenshot of my testcase on current Aurora 9.0a2 (2011-11-05)

Comment 15

[The 8472]

Banding issues could be solved via bug 627771

The odd color stops mentioned in comment 5 are a separate issue from banding.