255584 - Clicking on link in list causes item to shift position

Status: RESOLVED FIXED

(Core :: Layout: Block and Inline, defect)

Description

[Nicholas Allen]

FF 0813 trunk WinXP

Clicking on a link in a list causes the list to shift down on the screen and
does not activate the link.  Clicking on the link a second time works normally.
 It seems to take a combination of circumstances to hit this bug:

* link must be in the first item of the list
* list item has a text indent style
* page must be in standards mode

The type of list doesn't seem to matter.

This came from a real world page, I wasn't doing this just for fun.  It looks
like this broke in the last few weeks but I don't have a firm date.

Test case to follow.

Comment 1

[Nicholas Allen]

Attachment: Test case

Comment 2

[Nicholas Allen]

Broken back to at least 0727.  I don't have any older trunk builds.

Comment 3

[Mats Palmgren (inactive)]

Regression occured 2004-07-16-05 -- 2004-07-17-06.
In that span, bug 151375 looks like the most likely culprit to me.

Comment 4

[Jason Bassford]

Yes, I've been noticing this with the error pages for several weeks now...

Comment 5

[Nicholas Allen]

It looks like bug 151375 is the culprit.  This bug stops happening when the
focus outline is disabled with a user style.

Comment 6

[Mats Palmgren (inactive)]

Attachment: Minimal testcase

Comment 7

[Mats Palmgren (inactive)]

Attachment: Another testcase

A vertical scrollbar appears on the DIV when hovering the text.
(this did not occur before bug 151375)

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Duplicate of bug 252920?

Comment 9

[Mats Palmgren (inactive)]

Probably. The testcases here does not involve tables though.

Comment 10

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Sorry, I don't believe this is a duplicate. The underlying problem here is that
when content extends above or to the left of the origin of an overflow:auto
element, we add a scrollbar even though it's useless.

Comment 11

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: fix

nsBoxToBlockAdaptor and nsGfxScrollFrameInner do not currently handle
situations where there's overflow above or to the left of the origin. With this
patch, we just cut off that area so that you can't scroll there. It's not the
optimal solution but it's the simplest one. We can do a better fix after scroll
refactoring is done.

Comment 12

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Checked in.

Comment 13

[Bill Mason]

Please note that this fix is believed to have caused regressions in bug 256436
and probably bug 256465.

Comment 14

[Kenneth Herron]

Reopening - The patch was backed out; see bug 256436.

Comment 15

[Robert O'Callahan (:roc) (email my personal email if necessary)]

The nsGfxScrollFrame patch was bogus ... the scrolled view's bounds' origin is
the (negative) scroll position, so taking the scrolled view's bounds' YMost() as
the scrollable height is completely wrong when the scroll position is not (0,0).
So when we clicked on a link, it got focused, outline change caused a reflow
which recomputed the scrollable height to be something completely wrong, and
then scrolled the page up to try to avoid displaying any area below the
purported scrollable height.

However, I still believe that the nsBoxToBlockAdapator patch is correct. As I
describe here:
http://bugzilla.mozilla.org/show_bug.cgi?id=256436#c24
my tests show that it's doing the right thing.

I'm going to reland it.

Comment 16

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Relanded

Comment 17

[Anton van Bohemen]

Even though the test cases work fine, I still see this bug on this page:
http://white.sakura.ne.jp/~piro/xul/_tabextensions.html.en Try to click on
"download" for example, it shifts. Using Mozilla/5.0 (Windows; U; Windows NT
5.0; en-US; rv:1.8a3) Gecko/20040828 Firefox/0.9.1+ (BlueFyre)

Comment 18

[Jason Bassford]

Also, the "Try again" link from error pages jumps - I couldn't find another bug
on that one specifically and had hoped that this would fix it.

Comment 19

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Hmm. Please file new bugs for those tests. It would really help if you can
minimize the test cases.

Comment 20

[Jason Bassford]

I filed bug 257612 on the error page "Try again" problem.