User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7) Gecko/20040614 Firefox/0.8 Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7) Gecko/20040614 Firefox/0.8 Changing the font size causes some of the strings (at the left side of the page) to display incorrectly, meaning, the logical end of the string is flipped and is displayed from end to begining. It doesn't seem to happen on all hebrew encoded web pages. Reproducible: Always Steps to Reproduce: 1.Fire up one of the web pages (http://www.techno.co.il;http://www.start.co.il;...) 2.Change the font size (Ctrl + || Ctrl -) 3.Notice the strings on the left side of the screen. Actual Results: The logical end of the string is flipped and is displayed from end to begining. (On the left part of the page, it doen't happen for all the strings on the page). Expected Results: Display the string correctly, just as it was displayed when the page was loaded at the first time.

Can you provide a screenshot? This sounds similar to bug 141004.

Looks ok here with camino on osx 10.3.4: User-Agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.7) Gecko/20040623 Camino/0.8

It does seem to be related to bug 141004, but changing the font size gurantees you will be able to reproduce the bug (at least on Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7) Gecko/20040614 Firefox/0.8). However as you can see from the screenshot itself it does not happen with every string on the page, so there is a (very slight) possibility that the page will be altered in such a way that it will appear to be fine, just give it some time, it will show itself.

I wasn't able to reproduce it on OS X.

I can reproduce this in Windows, but not in BeOS. Prog.

This doesn't seem to occur with my older nsBidiPresUtil. I'll try to investigate further.

Bug is still here (Firefox v1.0 on Windwos platform). Notice that it is enough just to re/choose the character encoding from the view menu. This causes the page to display corretly. Hope it helps... (Some pages of websites like www.ynet.co.il are having this issue without the font resizing action).

Simon, thanks for the hint: this really accompanies RemoveBidiContinuation(). I'm still running into debugging problems (unable to inspect some objects; btw, does anyone have an idea what may be causing that?), but I believe my proposed small fix is right and has low regression risk. - Along with other vars, |fragmentLength| should be updated as well. This forces advancing to the right frame. Otherwise, the frame instance gets out of sync with the |lineOffset| (and other friends).

Shai, forgot to say that I could never reproduce that without resizing font. Please can you give an exact URL where it occurs (and maybe scenario)?

I have checked it further more, seems that the issue is very illusive. I could give you a link for a page that the shows only I have noticed that sometimes it is loaded properly and sometimes not. * Thought it had to do with different page encodings - the encodings were the same on both occurrences. * Thought it had to do with different pages I open the link from (e.g., more than one page directing me to the same problematic link) - but on both directing pages it appeared to happen randomly. * Thought it had something to do with tabbed browsing - but no. * Also tried to changed the encoding of the directing page but it made no difference. * Noticed that the encoding is in which the page is loaded is 'Hebrew (Windows 1255)', but as I said, re-choosing it fixes the problem. Never the less, here is a link that works "fine" for me: http://www.ynet.co.il/articles/0,7340,L-3035617,00.html Notice that is occurs after the 2000 figure, and changing font size always causes the bug to show (at least for me). Let me now if you managed to reproduce it. Shai

Comment on attachment 171415 [details] [diff] [review] patch sr=dbaron, although I don't pretend to understand what this code is doing

Patch checked in.

*** Bug 289798 has been marked as a duplicate of this bug. ***

Using version 1.0.3 of Firefox and that bug is still there. The quick eye had seen occasionaly that the string gets flipped on one of the renderring steps, that is, it is loaded and displayed properly when the page initially loads/rendered, and for some reason (one of the other objects on the page I guess) suddenly it gets flipped. Hope it helps.

Shai, this fix is not included in the 1.0.x branch (only 1.1+)