User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.2) Gecko/20070222 SeaMonkey/1.1.1 Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a6pre) Gecko/20070702 Minefield/3.0a6pre After Bug 9101 "(shy) – Break lines at soft hyphens (­)..." was reported to be fixed, I found the following behavior: Yes, ­ is working now. But the character ​ (zero width space) is rendered now incorrectly (which used to be fine before the fix). The line breaking behavior is correct, but a square is rendered for this character instead of just nothing. Reproducible: Always Steps to Reproduce: 1.TestText​TestText​TestText 2. 3.

The tests at the bottom of that page WFM on XPSP2.

(In reply to comment #1) > The tests at the bottom of that page WFM on XPSP2. Yes, works on Windows XP Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a6pre) Gecko/20070630 Minefield/3.0a6pre

Seems WFM on Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a6pre) Gecko/20070702 Minefield/3.0a6pre ID:2007070204 too

This is how ­ and ​ are rendered on my display.

This screenshot was taken on the same computer with firefox-2.0.0.4.

(In reply to comment #4) > This is how ­ and ​ are rendered on my display. Probably a problem with some fonts? http://www.quirksmode.org/quirksmode.css uses "georgia,times,serif". Your font doesn't look like georgia. Which font is used?

Please check your testcase in a build from, say, last week. I think you'll find that the soft hyphen changes I just landed were not responsible for this regression, that it happened earlier. This is probably a bad font. We can work around it, but I'd like to verify that it's a font issue first.

> Your font doesn't look like georgia. Which font is used? I created a simple HTML page with "TestText​TestText​TestText&#8203..." and tried different font settings in Firefox Preferences. Fonts are influencing the problem, however I see this problem with most of the fonts. With these fonts the problem remains, e.g. Adobe Courier, Adobe Helvetica, Adobe Times, Arial, Georgia, cmr10, SUSE Serif, Courier New, etc. A few fonts don't show anything wrong: .e.g. Courier With "Gnu Unifont", a small glyph saying "NUL" is displayed at the positions of ​. I tested the situation again with the binary release granparadiso/alpha5 (Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a5) Gecko/20070605 GranParadiso/3.0a5): It is as you expected: the problem with the squares at the 8203 positions is already visible in this version, and therefore does not seem to be related with the soft hyphen fix. I tested it also with granparadiso/alpha1, and already here I see the small squares.

OK. Does anyone else see the problem with those fonts? What version of Pango do you have?

(In reply to comment #9) > What version of Pango do you have? pango-1.8.1-4 SuSE Linux 9.3 (i586)

On OS X 10.4.10, ppc - and using the testcase mentioned in comment 8, I see what looks like a full space (U+0020) where the zero-width space is. This happens with most of the default OS X fonts, except Lucida Grande. That includes the MS fonts shipped with the OS (Arial, Trebuchet MS, Georgia, Times New Roman). Some mono-space fonts work correctly: Monaco, Courier. I can go back quite a long time - so far 'till the 20070524 Minefield build. (I double checked the test file using the latest WebKit build and Opera 9.2 - both where correct for all fonts).

I compiled different versions of Pango and tested them with a current build (changing LD_LIBRARY_PATH). I get broken "zero width space" display with Pango versions <= 1.10.4. I get correct "zero width space" display with Pango versions >= 1.12.0. (tested with glib-2.12.12) Firefox-2.0.0.4 always shows a correct display, independent of the version of Pango (tested with 1.8.1, 1.10.4, 1.12.0).

Thanks for that information, p.ernst! I just thought I'd point out that per the (tentative) Linux runtime requirements for Firefox 3, we'll require versions of Pango 1.14.x and greater. http://wiki.mozilla.org/Linux/Runtime_Requirements

This should be fixed, even for old Pangos, by the patch I just checked in (associated with bug 385423).

I'd make the container width little bigger, maybe 100em, just in case there are really bad fonts out there. Also the reference testcase is usually called "-ref.html". Other than that, these look good and could be put under layout/reftests/text. Thanks!