Régis Caspar Reporter
	Description • 19 years ago

This looks like a round error or something similar. I will attach a testcase.
When a CSS rule is used to set the text as Arial 14px, today nightly build has wrongly inter-letter spacing for this text. Additionally, when the text contains a link, there a huge space before and after this link.

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a1) Gecko/20060611 Minefield/3.0a1 ID:2006061104 [cairo]

Everything was OK on 20060609 for sure and I didn't notice the problem on 20060610 but I could have missed it.

Regression due to bug 340590???

	Régis Caspar Reporter
	Comment 4 • 19 years ago

From mozillazine (stilist): 
> Regis, can you add a comment to the spacing bug that links this image 
> [http://img85.imageshack.us/img85/7475/wikispacing2mu.gif] and mention 
> I'm seeing it on Wikipedia, but not in the testcase you gave?
> I'd do it myself, but it incorrectly gives me an error about trying to add 
> keywords (which I'm not)."

	Ria Klaassen (not reading all bugmail)
	Comment 5 • 19 years ago

I see this and I wonder why the font looks so different.

	Régis Caspar Reporter
	Comment 6 • 19 years ago

(In reply to comment #5)
> I see this and I wonder why the font looks so different.
I fond the problematic setting (font.minimum-size.x-western was set to 11)
Note FF needed a restart before taking the pref change into account.

	Régis Caspar Reporter
	Comment 7 • 19 years ago

Please disregard comment #6, sorry, with this setting reset (and thus not present in about:config) I see the problem.

	Peter van der Woude [:Peter6]
	Comment 8 • 19 years ago

Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9a1) Gecko/20060610 Minefield/3.0a1 ID:2006061013 [cairo]

WFM

I bet it's your fr-FR OS locale (again) Régis

	Régis Caspar Reporter
	Comment 9 • 19 years ago

(In reply to comment #8)
> I bet it's your fr-FR OS locale (again) Régis
I don't know, perhaps. Perhaps different Arial version (I've Asian text support installed)

FWIW, I have no such problem in my Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a1) Gecko/20060606 Firefox/3.0a1 (debug, home compiled)

	Régis Caspar Reporter
	Comment 10 • 19 years ago

Another information: Using Peter's testcase and my system default usp10.dll (v.1.420.2600.2180), the last paragraph is badly rendered (see http://img69.imageshack.us/img69/4935/screenshot0026tt.png). With another version of usp10.dll (v.1.473.4060.0, most recent here) which sits in IE6 directory, the third paragraph is badly rendered (see http://img147.imageshack.us/img147/969/screenshot0019xs.png).

	Régis Caspar Reporter
	Comment 11 • 19 years ago

I finally fixed this issue with font character inter-spacing (and en.wikipedia.org crashing Firefox) by using the most recent known version of "usp10.dll" (1.613.5291.0 see http://en.wikipedia.org/wiki/Uniscribe). 

Should I WORKSFORME this?

	Bill Gianopoulos [:WG9s]
	Comment 12 • 18 years ago

(In reply to comment #11)
> I finally fixed this issue with font character inter-spacing (and
> en.wikipedia.org crashing Firefox) by using the most recent known version of
> "usp10.dll" (1.613.5291.0 see http://en.wikipedia.org/wiki/Uniscribe). 
> 
> Should I WORKSFORME this?
> 

No.

For starters, if the problem is really with usp10.dll and not with Mozilla, the correct resolution would be INVALID and not WORKSFORME.

That said, in this case, Mozilla needs to work with the earliest platform that the browser is currently supported under which is Windows 2000.  The version of usp10.dll that comes with Windows 200 is 1.325.2195.1340.

So, in my opinion, this needs to either work correctly with usp10.dll version 1.325.2195.1340 and later or be modified to not depend on usp10.dll at all.  As I read from the Microsoft site, vendors are not licensed to package usp10.dll with their application, but may take advantage of versions of usp10.dll currently installed for whatever reason on the target system.  Since we have no control over what version of usp10.dll is currently installed we can only assume that it is at least 1.325.2195.1340 as that is the version that corresponds to Windows 2000.

	Stuart Parmenter Assignee
	Comment 13 • 18 years ago

if we're crashing due to an old uniscribe we should figure out why.  it kind of sucks that we can't redist newer versions and that some features will work better with newer versions, but thats where we are.

I'll try to see if I can get an older version on my system (vmware w2k install or something).

fwiw, I've been doing all my testing with 1.420.2600.2180 (WinXP sp2) and 1.613.5291.0 (MSVOLT 1.2)

	Stuart Parmenter Assignee
	Comment 14 • 18 years ago

fwiw, at least with the 2nd testcase, we don't actually call in to uniscribe so the version that you're using shouldn't matter at all.  Looking in to whats up -- I suspect its a difference between the ABC width and the extents width.

	Régis Caspar Reporter
	Comment 15 • 18 years ago

(In reply to comment #12-14)
Thanks Bill and Stuart for your answers.

Stuart, If you need some usp10.dll versions for testing I can provide you some.

	Stuart Parmenter Assignee
	Comment 16 • 18 years ago

also I should note that with a small hack and a bug fix if I force this testcase to go through the uniscribe code path rather than the "fast" path things render properly.  Need to figure out what I'm doing wrong in the fast path.

Also, the crash could be due to the bug fix above if for some weird reason you managed to actually get your build to go in to uniscribe for the testcase currently.

	Régis Caspar Reporter
	Comment 17 • 18 years ago

(In reply to comment #16)
> Also, the crash could be due to the bug fix above if for some weird reason you
> managed to actually get your build to go in to uniscribe for the testcase
> currently.
If you are talking about the Wikipedia crash, it was probably related to bug 341865 because the crashing part was in the bottom of the page ("Wikipedia languages") and was the Thaï part.

	u88484
	Comment 18 • 18 years ago

I seen this yesterday on news.google.com, looked like the 4th paragraph in the testcase.

	David Clark
	Comment 19 • 18 years ago

I have found that the font weridness in the last paragraph of the testcase goes away if you disable the Supplemental language support (though I haven't tested if both selections need to be disabled).

	Stuart Parmenter Assignee
	Comment 20 • 18 years ago

this seems to be fixed..?

	Régis Caspar Reporter
	Comment 21 • 18 years ago

Seems fixed here too (with usp.dll v.1.420.2600.2180)

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a4pre) Gecko/20070403 Minefield/3.0a4pre ID:2007040304 [cairo]