Bug 4824 – CSS "border-bottom-style: dotted" attribute sometimes draws extra pixel

• Status: REOPENED
• Component: Layout and Rendering

[John Sullivan]

This bug is also in Radar as <rdar://4045083>

The CSS border-bottom-style:dotted attribute sometimes draws an extra pixel in the bottom border when 
the link length is a certain length.  See screenshot.  For "Fire" and "Colloquy" the bottome borders look 
fine;  but for "LimeWire" and "Fractality" you can see an extra pixel at the right side of the border;  almost 
like an ink smudge.

What's probably happening is that it's always drawing a last pixel at the end of the bounds... and this 
sometimes happens to line up with the dotted pattern, sometimes not.

(In reply to comment #0)
> What's probably happening is that it's always drawing a last pixel at the end of the bounds... and this 
> sometimes happens to line up with the dotted pattern, sometimes not.

This is intended (QWKPainter.mm:295):
        // Do a rect fill of our endpoints.  This ensures we always have the
        // appearance of being a border.  We then draw the actual dotted/dashed line.

ideally it should be done only if two borders meet at the given endpoint.

[Ahmad Saleem]

(In reply to mitz from comment #1)
> (In reply to comment #0)
> > What's probably happening is that it's always drawing a last pixel at the end of the bounds... and this 
> > sometimes happens to line up with the dotted pattern, sometimes not.
> 
> This is intended (QWKPainter.mm:295):
>         // Do a rect fill of our endpoints.  This ensures we always have the
>         // appearance of being a border.  We then draw the actual
> dotted/dashed line.
> 
> ideally it should be done only if two borders meet at the given endpoint.

I tried to search for this comment in WebKit source using Searchfox.org (easier to use when trying to search comment) but I couldn't find this comment.

I am not sure on the radar bug, whether we have any testcase etc.

Can we close this now?

Adding @ap - to help on Radar bit.

[alan]

I am not seeing any test case on this. Not sure what the right category for "needs more information" in bugzilla but I am somewhat convinced we already addressed this so -> config changed.

[Alexey Proskuryakov]

Created attachment 466261 [details]
test case

[Alexey Proskuryakov]

Doesn't seem fixed. All browser engines are different here, but at least on this test case, Firefox seems nicest.

[alan]

(In reply to Alexey Proskuryakov from comment #4)
> Created attachment 466261 [details]
> test case
👍

[Ahmad Saleem]

@Alan & Alexey - I think bug 268448 might have progressed it slightly. Since Safari 17.5 is different from Minibrowser.

https://commits.webkit.org/276585@main