Re: [css3-color] [css3-background] translucent borders in Webkit

On 02/25/2011 03:40 PM, Brad Kemper wrote:
>
>
> On Feb 25, 2011, at 8:36 AM, Simon Fraser<smfr@me.com>  wrote:
>
>> On Feb 24, 2011, at 11:24 pm, Brad Kemper wrote:
>>
>>> I've recently noticed that when using rgba borders in webkit,
>>> in which the alpha is less than one, that the corners get double
>>> amounts of "ink" at the corners, as though each adjoining border
>>> edge overlapped the others at the corners. It is very unwanted,
>>> always, and the other browsers I checked (Firefox, IE9) don't do
>>> that, but I can't find anything in CSS3 Color or Backgrounds&
>>> Borders 3 that prohibits it. I think it should.
>>
>> It's a bug in WebKit<https://bugs.webkit.org/show_bug.cgi?id=21835>
>>
>> Simon
>
> OK, that's good. Should it be in the spec too, so there can be a test for it?
> Something like "when two adjoining borders of like color meet at a corner,
> the effect should be as a border with continuous color flowing around the
> corner, and not of overlapping lines."?

I don't think that explaining how not to do graphical joins is the job
of a CSS spec. This is obviously a bug.

If there's someone genuinely arguing that it is not a bug, then argue
that the spec says to use the specified color for the border regardless
of what graphical process underlies the border-painting, and overlapping
two regions of rgba does not result in the border having that specified
color. It is therefore nonconformant. I don't think we need an explicit
statement about this particular variant of nonconformance.

"Also, the UA must not crash while processing any pages containing CSS."

~fantasai