bugFree UCS Outline Fonts - Bugs: bug #39474, Kerning too narrow between...

 
 

bug #39474: Kerning too narrow between "y" and "a"?

Submitter:  Rosella Capriotti <rosy58>
Submitted:  Fri 12 Jul 2013 07:07:03 PM UTC
   
 
Category:  individual character(s) Severity:  3 - Normal
Item Group:  letter positioning poor Status:  Fix posted
Privacy:  Public Assigned to:  Stevan_White
Open/Closed:  Open Release:  * 2012-05-03
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Mon 25 May 2015 01:23:11 PM UTC, comment #15: 

I finally found the problem.  It was simply my mistake.
In a separate table meant to kern letters with punctuation (especially extreme cases such as y with comma, etc) the table was also kerning letters with other letters -- the cumulative effect of kerning was thus doubled.

A fix is now in SVN for Serif.

Steve White <Stevan_White>
Group administrator
Mon 15 Jul 2013 10:44:55 AM UTC, comment #14: 

Rosella,

This is not about the kerning values, or particular kerning pairs.
It appears to be an overall doubling of the kerning amount.
The glyphs 'y' and 'v' only show it a lot because they are kerned more than most letters (due to their extreme shapes).

As I said, something in the building of the font is triggering (what I think is) a bug.
I can make it go on and off depending on how I build the font, but I don't know exactly what it is yet.

I'll let you know!

Steve White <Stevan_White>
Group administrator
Mon 15 Jul 2013 10:32:23 AM UTC, comment #13: 

Hello, what I said about "y" applies to "v" as well, I've just remarked it while I was writing, do you want me to provide you with a sample?

Rosella Capriotti <rosy58>
Sun 14 Jul 2013 06:28:35 PM UTC, comment #12: 

Hi,

As I understand it, your PDF file shows kerning working correctly.

I'm still haven't narrowed down what triggers the "double-kerning" behavior.

Perhaps you should open a separate report concerning ligatures.
(However, please provide a test file, and specify all relevant version numbers.)

Cheers!

Steve White <Stevan_White>
Group administrator
Sun 14 Jul 2013 05:20:24 PM UTC, comment #11: 

Hello, it looks FreeSerif kerns all right in XeLaTeX, this is 12pt text. This things makes me think LibreOffice causes this mess, remember LibreOffice also ignores basic ligatures. Maybe you should tune up the implementation to suit Microsoft Word oddities (you told me LibreOffice emulates Word's behaviour).

(file #28570)

Rosella Capriotti <rosy58>
Sun 14 Jul 2013 12:35:33 PM UTC, comment #10: 

Oops... please delete the previous comment, it wasn't meant to be here :(

Rosella Capriotti <rosy58>
Sun 14 Jul 2013 12:34:12 PM UTC, comment #9: 

By the way, I have one question about the activation of this glyph: FontForge says it's activated through salt, no mention of locl, but does it mean it will appear even in a Russian text, if one turns salt on? This is inappropriate, the glyph should ideally be the combination of locl+salt not just salt alone. And what's more, I suggest you make use of stylistic sets, just as Alexey does (you remember the manual of his Greek fonts?). I see you have implemented it rather inconsistently for Italic Serbian "de" alone, but I can't to understand why you did it and why for this glyph alone. My take is that stylistic set should be implemented to turn on "sha", but on the understanding that it will do so only if Language=Serbian.

Rosella Capriotti <rosy58>
Sun 14 Jul 2013 10:55:47 AM UTC, comment #8: 

I'm going to provide you with a sample text as soon as possible, about the ligatures I was referring to default stuff like "fi", which is the most visible one. But they should be turned on by default, if they're available, according to the LibreOffice manual. And other fonts' ligatures do work in LibreOffice, like MinionPro, OldStandard TT (Alexey's font), Linux Libertine G and so on.

Rosella Capriotti <rosy58>
Sun 14 Jul 2013 09:58:56 AM UTC, comment #7: 

Hi,

Well, have a look.  But I think I'm seeing the problem in other apps -- I can turn it off and on, so I should be able to track the trigger down.

Some font features are meant to be active by default ( 'rlig' and 'liga') while others are activated by the application or user ('hlig').

Which ligatures are you referring to?

Recently, the main font rendering libraries in Linux have changed: it used to be Pango, now it's Harfbuzz.  Firefox and XeLaTeX both use Harfbuzz now.  LibreOffice uses its own, called ICU.  These do a lot of things differently.  (Google Chrome and Opera also have their own font rendering libraries.)

Also, LibreOffice often emulates MSWord's behaviour--for better or worse.  Word is also no genius when it comes to modern font features.  In fact, in Windows, web browsers and simple editors such as WordPad often do a better job than Word, in their use of finer font features.

Steve White <Stevan_White>
Group administrator
Sat 13 Jul 2013 08:22:00 PM UTC, comment #6: 

OK, I see. Do you want me to check how it works out in XeLaTeX? Maybe it could be useful to you.
Please satisfy my curiosity: why do FreeSerif's ligatures work only in XeLaTeX? I can't get them to work in LibreOffice but neither had I activated any special feature in XeLaTeX, they just popped up unexpected, I used to believe FreeSerif didn't make allowance for them, actually!

Rosella Capriotti <rosy58>
Sat 13 Jul 2013 08:07:38 PM UTC, comment #5: 

Funny.  I thought I didn't see it in L/O. 
I see it in Firefox and Gedit though.
The effect looks like the kerning values are doubled (applied twice?)

So far, have eliminated a couple of possibilities.

  • It isn't a duplicate kerning table entry: removed all GPOS tables except one--no effect.
  • It isn't triggered by enabling "old style kern" table.


If I generate the font directly from a running Fontforge session, the letters are kerned as expected.  If I use the "make" script, the bad effect occurs.

Attached are some images of good and bad in Firefox.


Steve White <Stevan_White>
Group administrator
Sat 13 Jul 2013 09:32:37 AM UTC, comment #4: 

Hello Steve,
Indeed I noticed this glitch in LibreOffice with FreeSerif 12pt, here's the screenshot with 200pt.


Rosella Capriotti <rosy58>
Sat 13 Jul 2013 12:02:39 AM UTC, comment #3: 

No, I was looking at the tables cross-eyed.
I still see the funny kerning in Firefox.
But I removed all the kerning tables from FreeSerif except the one, and this still happens.

I still don't know what application you're using.

But try copying your text into an OpenOffice document, and making the text a large FreeSerif.  See what happens.

Steve White <Stevan_White>
Group administrator
Fri 12 Jul 2013 11:09:28 PM UTC, comment #2: 

Hi

I see it, and I figured it out.

In Firefox, text with those pairs does seem terribly over-kerned.
Yet the kern table I looked at shows them lightly kerned.
(ran through several obvious explanations...ended up with: )

Ah.  There are stray entries for these letters in another table, a special purpose one meant for combinations with languages that use the IPA for writing.

Just a goof-up on my part -- should be easy to fix.

Thanks for pointing this out!

Steve White <Stevan_White>
Group administrator
Fri 12 Jul 2013 10:53:27 PM UTC, comment #1: 

Hi again,

The exact appearance of rendered text depends greatly on the software used to render it, as well as other parameters in the environment.

Please provide images, and report the specific applications you use to render the text.

I just looked at 'ya' 'ye' and 'yo'.  In each case, the distance between any part of the letters is greater than the stem width, which has been one of my rules of thumb.  The kerning only takes these round letters up to the farthest extent of the top right serif on the 'y', so they are not actually under the letter.  So... to me it looks right. 

I'll have to see your screenshot.

Steve White <Stevan_White>
Group administrator
Fri 12 Jul 2013 07:07:03 PM UTC, original submission:  

Hello Steve,
I was wondering whether the kerning in pairs like "ya" (Latin alphabet this time, no Cyrillic) is too narrow, the letters almost collide. Also, what about pairs like "ye" and "yo"? Aren't they too narrow too? I remember when we were discussing Cyrillic kerning, you told me letters are like enclosed in an ideal box and the lines circumscribing those boxes mustn't collide when dealing with kerning. But in cases like "ye" and "yo" they actually do so, is "y" an exception in typography? Let me know, I'm curious :)
Regards,

Rosella Capriotti <rosy58>

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:
   
   
Comment:
   

Attached Files
file #28570:  Kerning of Y.pdf added by rosy58 (3KiB - application/pdf)
file #28563:  yoga-ff-make.png added by Stevan_White (41KiB - image/png - font generated from makefile shows excessive kerning; the other -- OK.)
file #28564:  yoga-ff-o-s-kerns.png added by Stevan_White (41KiB - image/png - font generated from makefile shows excessive kerning; the other -- OK.)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by Stevan_White (Posted a comment)
  • -email is unavailable- added by rosy58 (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only logged-in users can vote.

     

    Follow 9 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2015-05-25 Stevan_White StatusNeed info Fix posted
    2013-07-14 rosy58 Attached File- Added Kerning of Y.pdf, #28570
    2013-07-13 Stevan_White Attached File- Added yoga-ff-make.png, #28563
        Attached File- Added yoga-ff-o-s-kerns.png, #28564
    2013-07-13 rosy58 Attached File- Added Capture du 2013-07-13 11:31:21.png, #28558
    2013-07-13 Stevan_White StatusProceeding Need info
    2013-07-12 Stevan_White StatusNeed info Proceeding
    2013-07-12 Stevan_White Assigned toNone Stevan_White
    2013-07-12 Stevan_White StatusNone Need info

    Back to the top

    Powered by Savane 3.13-f8d8.
    Corresponding source code