162841 - radio +   + href will wrap

Status: NEW

(Core :: Layout: Text and Fonts, defect, P3)

Description

[Ryan Gustafson]

A radio followed by a   followed by an href will wrap.  However, a radio
followed by a   followed by text will not.  I'm attaching test case showing
the problem.  Works fine in IE.  In Mozilla 1.0 though, the top table (fixed
width) does a break, and the bottom table (not fixed width) will break if you
resize your browser window small enough.

Comment 1

[Ryan Gustafson]

Attachment: Two tables, 1st fixed width, 2nd not. both can show the bug

Resize your browser smaller to get the 2nd table to show the problem.

I'm not reporting this as a bug, but looks like the text itself will draw
_outside_ the bounds of the table as well.

Comment 2

[Wesha]

Problem confirmed 2002081504/winXP.

Reporter: Please ALWAYS tell us your BuildID. Also, it is highly recommended
to use Bugzilla Helper, not Advanced Entry form for reporting bugs.

Comment 3

[Ryan Gustafson]

Apologies, didn't realize I used the advanced, was wondering why it was not as
simple/friendly as the last time I submitted a bug.

Build ID is 2002052306, would be 1.0-pre3, I just realized.

Just downloaded 1.0-release, Build ID is 2002053012, verified to exist there as
well.

Comment 4

[Boris Zbarsky [:bzbarsky]]

To intl; this is a line-breaking problem.

Comment 5

[Shanjian Li]

take it myself.

Comment 6

[Shanjian Li]

accept.

Comment 7

[Shanjian Li]

I could not reproduce the problem. Would you mind try it on latest trunk build
and attach a screen shot. I will update my local tree and try again.

Comment 8

[Ian Neal]

Is this anyway related to bug 122441 or bug 139723?

Comment 9

[Shanjian Li]

This bug looks like a regression. I could not reproduce the bug before, but I
could not. And it seems ok to me with Netscape7.0.  As I looked into the
problem, I found that we did things differently with normal text run and
connecting frame text run. 
I believe this bug reveals several problems. nsLineLayout::LineIsBreakable and
nsTextFrame::MeasureText are 2 of them. 

Yuying, could you help figure out when do we began to have this regression? 

Comment 10

[Yuying Long]

This seeems not a regression for me, cause I don't see there is different
between N6.23 and N7.0.

However, there is a difference when set page encoding as Chinese (TC or SC) and
other encodings (western, Japanese, Korean...etc).
With Chinese encoding, there is no line wrap in 1st table while with other
encodings will get line wrapped.

Comment 11

[Shanjian Li]

thanks yuying for checking this for me. 

Here is my guess of what's happenning:
When there is frame before the text frame, "mCanBreakBefore" is set to true.
This value is ignored for text measuring in same text frame when mX == 0. For
joining frames, mCanBreakBefore is observed and we break if the word is too
long. Clearly, there is a inconsistency here. The inconsistency is easy to fix,
but the bug reveals a more complicated problem. That is the text extruding
outside table cell boundary. The reason is we use maxWordWidth to computer the
width of table cell. This computation will not work if the longest word is not
breakable with a non-text frame, eg. radio buttion in the bug report. This is a
rather troublesome problem to handle. 
 
change component to layout, this is not line-breaker problem.

Comment 12

[Ian Neal]

Really reassigning

Comment 13

[Boris Zbarsky [:bzbarsky]]

.