617239 - min-intrinsic width computation ignores soft hyphens

Status: RESOLVED DUPLICATE of bug 418975

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

Description

[Ilguiz Latypov]

User-Agent:       Mozilla/5.0 (Windows NT 5.1; rv:2.0b8pre) Gecko/20101206 Firefox/4.0b8pre
Build Identifier: Fresh nightly 4.0b8pre

When I load the attached HTML file into my browser, the browser appears unable to break the long string starting with "boost" around "struct boost" and ending with "filesystem::" around "boost:filesystem:basic_path".

The width of the inner window at which a horizontal scroll bar appears is 1423.


Reproducible: Always

Steps to Reproduce:
1. Open break.html in the browser.
2. Resize the browser window so that the margin attempts to wrap the end of the string "boost::filesystem" in the string "boost::filesystem::basic_path".

Actual Results:  
The browser fails to break the string.  The browser shows a horizontal scroll bar.


Expected Results:  
The browser should re-flow the text so that the first half of the line breaks by one of the numerous soft hyphens.


HTML 4.01 Strict.

Comment 1

[Ilguiz Latypov]

Attachment: a test of line breaking by soft hyphens

Comment 2

[Ilguiz Latypov]

Attachment: a test of line breaking by zero width spaces

Comment 3

[Ilguiz Latypov]

Replacing soft hyphens ­ with zero width spaces ​ lets the browser wrap the line.  (As expected, the browser will not show hyphens.)

Comment 4

[Boris Zbarsky [:bzbarsky]]

Yeah, this is sorta odd.  Masayuki-san, do you know what's going on here?

Comment 5

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

I have no idea. If the test is out of table cell, the line breaker works fine. So, isn't this a bug of table layout?

Comment 6

[Boris Zbarsky [:bzbarsky]]

Attachment: Testcase not using tables

This seems to be a general problem with shrink-wrapping; in particular the min width that's computed is too wide.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Looks like nsTextFrame::AddInlineMinWidthForFlow doesn't treat ­ as a line-break opportunity.  There's a comment like so:

  // XXX Should we consider hyphenation here?

Seems like we should.

Comment 8

[Jonathan Kew [:jfkthame]]

This looks a lot like bug 418975, and the patch there should fix it.