290567 - regression: most hebrew content layed out in reverse

Status: RESOLVED FIXED

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

Description

[Eyal Rozenberg]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050128 Mnenhy/0.7
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050128 Mnenhy/0.7

With a build from yesterday, many pages in Hebrew have some of the text layed
out in the proper order: TFEL OT THGIR MORF, but the bulk of the text is layed
out FROM LEFT TO RIGHT - the wrong order for Hebrew (note I'm not talking about
paragraph alignment). Plus, the layout is kind of 'messed up' and letters
overlap. This does not happened with a build from about 2 weeks ago. Also,
removing all style from the page (Edit|Use Style|None) fixes the problem.

Example URL:

http://www.earendil.ath.cx/index.shtml.he

This may or may not be related to bug 289798.

Reproducible: Always

Steps to Reproduce:

Comment 1

[Asaf Romano (gone)]

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b2) Gecko/20050415
Firefox/1.0+

wee, probably a regression from bug 284927.

Comment 2

[nyh]

*** Bug 290627 has been marked as a duplicate of this bug. ***

Comment 3

[nyh]

Another example is: http://nadav.harel.org.il/hindex.html (my homepage)

Note how the headers are shown correctly (Hebrew letters flowing from right to
left), but the paragraph is incorrect - letters flowing from left to right.

I too traced the cause of this is bug to a style: in this case, the "text-align:
justify;" style. When this style is removed from the paragraph, it returns to
being correctly shown.

By the way, when you select text in the broken paragraph, the text jumps around.
So even if you don't know Hebrew, you can see something is broken.

This is indeed a regression. This problem did not exist in earlier Mozillas - my
homepage always looked correct until very recently (I don't remember which
version broke it).

Comment 4

[Asaf Romano (gone)]

Not reproducanle on:
Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.8b2) Gecko/20050416
Firefox/1.0+

(i.e. os w/o native bidi support).

Comment 5

[nyh]

I reproduced it in the latest Firefox 1.0.2 that comes with Fedora Core 3
updates. Maybe this bug is related to Pango support, and this should be marked a
dup of bug 284927 ? (but someone reproduced this on Windows - does Windows also
use Pango?)

Comment 6

[Eyal Rozenberg]

(In reply to comment #5)
> Maybe this bug is related to Pango support, and this should be marked a
> dup of bug 284927 ?

It's not a dupe, it's a regression caused by the fix for 284927, on which it
depends.

> (but someone reproduced this on Windows - does Windows also use Pango?

Indeed, it does not, but it seems somehow the Mozilla code now treats justified
RTL-language content as though Windows is Pango-like, and pre-flips the order
for it - which it shouldn't.

Comment 7

[Simon Montagu :smontagu]

Attachment: Patch

Comment 8

[Mike Kaply [:mkaply]]

Comment on attachment 180946 [details] [diff] [review]
Patch

preapproving
a=mkaply

Comment 9

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

Comment on attachment 180946 [details] [diff] [review]
Patch

d'oh! I should have spotted that.

Comment 10

[Simon Montagu :smontagu]

Checked in. My apologies for this stupid mistake.

Comment 11

[Caleb]

Fixed?

It looks like this patch _BROKE_ it.

I've had no problems before this (although I have spotted a few pages having
some content in reverse, but not overlapping text).

With this patch checked in I started seeing overlapping hebrew text.

Check this page: http://www.ynet.co.il/articles/0,7340,L-3074316,00.html
Scroll to the bottom (the talkback area). The parts that have some hebrew text
(the name of the author of the talkback) and date are totally messed up!

Attaching screenshot.

Comment 12

[Caleb]

Attachment: overlapping hebrew text

problematic areas underlined in red.

Comment 13

[Simon Montagu :smontagu]

(In reply to comment #11)
> With this patch checked in I started seeing overlapping hebrew text.

Please file a new bug on this problem and assign it to me.

Comment 14

[Simon Montagu :smontagu]

By the way, I'm fairly sure the regression on ynet.co.il was not caused by this
checkin. I don't know yet what did cause it.

Comment 15

[Eyal Rozenberg]

> Please file a new bug on this problem and assign it to me.

All are welcome to enjoy the lovely new bug 291188.