355917 - Lot of Unicode symbols missing in trunk

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Ria Klaassen (not reading all bugmail)]

I'm missing a lot of symbols on this page: http://www.alanwood.net/unicode/mathematical_operators.html

Regression range 1.9a1_2006092914 - 1.9a1_2006092916:
http://bonsai.mozilla.org/cvsquery.cgi?module=PhoenixTinderbox&date=explicit&mindate=2006-09-29+13%3A00&maxdate=2006-09-29+17%3A00

Caused by Bug 354887 ? or Bug 354966 ?

Comment 1

[Justin Kerk]

I assume this is the same bug as the hyphen in "blocking‑aviary" no longer being displayed at all? Is the regression window the same?

Comment 2

[Stuart Parmenter]

WFM.  354887 could be the problem, but that fixes a pretty ugly bug that causes us to look at way more fonts than we should for font selection.

please run your build with NSPR_LOG_MODULES=winfonts:5 NSPR_LOG_FILE=c:\foo and attach the output from loading that page.

thanks

Comment 3

[Ria Klaassen (not reading all bugmail)]

Attachment: log

Still seeing the same.

Comment 4

[Justin Kerk]

Attachment: log

Here's my log, I have a bunch of startup tabs so I trimmed it to the section that corresponded to the previous log, hope that's OK.

I do see the bug, in case that's not clear.

Comment 5

[Stuart Parmenter]

ria: are you on vista or just using vista fonts (Constantia)?  I know we've seen some issues reported with people using vista fonts on XP...

Comment 6

[Stuart Parmenter]

(In reply to comment #4)
> Created an attachment (id=242703) [edit]
> log

Can you check what version of AR PL ZenKai Uni you're using?  control panel/fonts/double click on the font.

Comment 7

[Justin Kerk]

(In reply to comment #6)
> Can you check what version of AR PL ZenKai Uni you're using?  control
> panel/fonts/double click on the font.

Version 0.1.20060903

And my build ID, because I forgot to post it before:
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a1) Gecko/20061018 Minefield/3.0a1 ID:2006101804 [cairo]

Comment 8

[Stuart Parmenter]

(In reply to comment #7)
> (In reply to comment #6)
> > Can you check what version of AR PL ZenKai Uni you're using?  control
> > panel/fonts/double click on the font.
> 
> Version 0.1.20060903

OK, I'll retest with that version.  I'm having trouble getting that font to be selected over some others.  I may need to add some more debugging output to get more useful logs to track this down further.

Comment 9

[Stuart Parmenter]

Attachment: possible fix

i'm a little afraid this patch will break some languages that might use zero-width glyphs on purpose (I can't think of any that would use glyphs (I can think of some that would use zero-width characters)).  Can either of you do a build with this?  If not I'll check it in tomorrow and we'll look out for regressions.

Comment 10

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

Comment on attachment 242715 [details] [diff] [review]
possible fix

> +            else if (mAttr[i].fZeroWidth == PR_TRUE) {

fZeroWidth is a WORD variable, not PRBool.
I think that you should use |if (mAttr[i].fZeroWidth)|

Comment 11

[Ria Klaassen (not reading all bugmail)]

(In reply to comment #5)
> ria: are you on vista or just using vista fonts (Constantia)?  I know we've
> seen some issues reported with people using vista fonts on XP...
>
Using Windows XP. Before 29 September 2006 I don't see the problem with the same font installed. I became aware of the problem when I missed a little symbol before all my mailto-links. 

Comment 12

[Ria Klaassen (not reading all bugmail)]

Attachment: trouble causing font

When I delete this font the symbol comes back

Comment 13

[Justin Kerk]

I don't have a build environment for Firefox, I plan on correcting that one of these days though.

Comment 14

[Stuart Parmenter]

i checked in this patch -- can you guys see if this fixes it for you?

Comment 15

[Ria Klaassen (not reading all bugmail)]

Yes, it is fixed now.

Comment 16

[Justin Kerk]

Yeah, that did the trick. The hyphen in blocking‑aviary is back now too.

Comment 17

[Simon Montagu :smontagu]

This totally broke rendering of scripts using zero-width joiners and
non-joiners. I'm not sure why there are still zero-width glyphs after calling
Shape(), but apparently there are.