Open Bug 1289828 Opened 6 years ago Updated 5 years ago

Dotted border doesn't completely redraw when its width changes


(Core :: Layout, defect, P3)

50 Branch



Tracking Status
firefox50 --- affected


(Reporter: arni2033, Unassigned)




(5 files, 1 obsolete file)

>>>   My Info:   Win7_64, Nightly 50, 32bit, ID 20160714030208 (2016-07-14)
1. Open url [1]
2. Wait 7 seconds

AR:  Border leaves small black areas on the page
ER:  No graphic artifacts

> [1] data:text/html,<div></div><style>div{border-bottom: 230px dotted black;}</style><script>for (i=230;i>=1;i--){(function(){var I=i;setTimeout(function(){document.querySelector("div").style.borderBottomWidth=I;},(230-I)*30)})() }</script>

 I marked this bug as blocking bug 382721, because this exact glitch didn't happen before that bug.
 However, there was a much worse bug that was fixed by that bug. Still it seems like a minor flaw
 in a completely new piece of code, so I decided to wait for response from developers.
Doesn't happen for me, on the following environments (I don't have win7 64bit env...):
  * Windows xp  32bit, Nightly 50.0a1 (2016-07-27) 32bit
  * Windows 7   32bit, Nightly 50.0a1 (2016-07-27) 32bit
  * Windows 8.1 32bit, Nightly 50.0a1 (2016-07-27) 32bit
  * Windows 10  64bit, Nightly 50.0a1 (2016-07-27) 32bit
  * Windows 10  64bit, Nightly 50.0a1 (2016-07-27) 64bit
  * OSX 10.11   64bit, Nightly 50.0a1 (2016-07-27) 32bit
  * OSX 10.11   64bit, Nightly 50.0a1 (2016-07-27) 64bit
  * Xubuntu     32bit, Nightly 50.0a1 (2016-07-27) 32bit
  * Debian      64bit, Nightly 50.0a1 (2016-07-27) 64bit
  * Android     x86,   Nightly 50.0a1 (2016-07-27) x86

Does it still happen on today's nightly with clean profile?
Also, does disabling hardware acceleration help?
Yes, disabling HWA helps.
Attached file testcase
Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:50.0) Gecko/20100101 Firefox/50.0

Confirming bug on win7 with HWA on only
Something specific to the version of D2D or D3D that is being used maybe.
mmm, I don't have any machine or VM that D2D can be enabled :/

anyway, as it's border-bottom, so that glitch could happen if the dot (circle) was rendered outside of the border area.
is there any case that accelerated PathBuilder::Arc+Fill rendering can become bigger...?
What I can think of are these 2 cases, the dot gets misaligned, or rendered bigger.
but I have no idea how that could happen only for the specific dot, not all dots.
Attached file testcase 2 (obsolete) —
Slightly better testcase, seems a bit more reproducible.
It also depends on window width for me.
Attached file testcase 2 (corrected)
Attachment #8775989 - Attachment is obsolete: true
See Also: → 1327888
Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.