418574 - List items can be misplaced if <ol> "start" or <li> "value" attributes change

Status: RESOLVED DUPLICATE of bug 203727

(Core :: Layout, defect, P2)

Description

[Jesse Ruderman]

The testcase and reference should look the same.  refdyn found this bug thanks to layout/reftests/bidi/413928-2.html.

Comment 1

[Jesse Ruderman]

Attachment: testcase

Comment 2

[Jesse Ruderman]

Attachment: reference

Comment 3

[Masahiro YAMADA]

Attachment: Screensho on windows

At WinXP, at first drawing, testcase and reference is NOT look the same.
But, After changing width of window, both looks the same.

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b4pre) Gecko/2008021904 Firefox/3.0b4pre

Comment 4

[Uri Bernstein]

Based on the regression date (2006-12-08), this is a regression from reflow refactoring.

Comment 5

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

layout regression, should probably block

Comment 6

[Simon Montagu :smontagu]

Nothing really bidi about this, except that the testcase has dir="rtl". There are cases of the same bug in ltr. 

Comment 7

[Jesse Ruderman]

Do you have a testcase that demonstrates the bug without dir="rtl"?

Comment 8

[Simon Montagu :smontagu]

Attachment: ltr testcase

Comment 9

[Simon Montagu :smontagu]

Attachment: Reference for ltr testcase

Comment 10

[Simon Montagu :smontagu]

From IRC:
<smontagu> i think i have a oneline fix for it
           namely to do FrameNeedsReflow instead of Invalidate at http://mxr.mozilla.org/seamonkey/source/layout/generic/nsBlockFrame.cpp#6522
           does that make sense?
<roc>      i guess ... we changed the number, so the frame size needs to change?
           except you can't call FrameNeedsReflow during reflow
           you need to do it as a post-reflow callback
           not sure what that would do to Tp
<smontagu> I don't think this is during reflow
<roc>      sure is
           http://mxr.mozilla.org/seamonkey/source/layout/generic/nsBlockFrame.cpp#915
           I wonder why we don't renumber lists after restyle/frame construction
           ask bz
           or dbaron
<smontagu> it also gets called from AttributeChanged
           which is the one that the testcase hits
<roc>      okay but you can probably cook up a testcase that hits it during reflow
           e.g. set things up so removing an element from the tree causes the bullets to renumber the same way
           or adding one

Comment 11

[Simon Montagu :smontagu]

Attachment: testcase for renumbering during reflow

Comment 12

[Simon Montagu :smontagu]

Attachment: reference for reflow testcase

Comment 13

[Simon Montagu :smontagu]

For the record, only the first testcase is a regression. The others have been broken since forever (tested in a build from 20010131).

Comment 14

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

Yeah, doing list numbering during reflow sucks.  My long term plan is to switch lists to using counters, which are processed at frame construction time.  I think there's even some work-in-progress lying around in a bug somewhere, but it's too big a project for 1.9 at this point.

Comment 16

[Simon Montagu :smontagu]

Can anyone still reproduce this on mozilla-central trunk? All the testcases seem to work for me.

Comment 18

[Simon Montagu :smontagu]

Attachment: reftest diff

Attachment 305283 [details] is equivalent to layout/reftests/bugs/203727.html, but I think the other ones are sufficiently different that it's worth checking them in.

Comment 19

[Simon Montagu :smontagu]

Checked in reftests to trunk as http://hg.mozilla.org/mozilla-central/rev/2c8799d7166e

Comment 20

[Simon Montagu :smontagu]

Checked in reftests to 1.9.1 as http://hg.mozilla.org/releases/mozilla-1.9.1/rev/dfbdf5cc37e7

Comment 21

[Tony Chung [:tchung]]

marking verified1.9.1 on Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.1b3pre) Gecko/20090130 Shiretoko/3.1b3pre Ubiquity/0.1.5.   

However, im a little perplexed why the fixed1.9.1 keyword was added here, and not in bug 418574 instead.  Since this was marked as a dupe of it.