[ft-devel] Font Validator 2.0, and libre font test results

[Hin-Tak Leung]

Hi,

It is a bit like the Mac OS X native binaries - it took a long time to get the 
command-line to work, then another week later, the GUI follows. So last weekend 
the rasterization test (which tests for truetype hinting instructions) which 
Microsoft did not release, got the beginning of an re-implementation in the 
Font Validator. That was the final missing piece that Microsoft did not 
release. Since then, I have tested against all(?) the fonts on Fedora linux 
(about 1750 of those), MS Win 8.1 (~300), Apple's (~700) and my private stash 
of just over 2000 rather old and buggy mostly CJK fonts - CJK fonts are few and 
far between, so you hang on to the vaguely useable and interesting ones, 
however buggy they are... It has been a week of repeated testing on almost 5000 
fonts in total.
  
Another week later, I am comfortable to call the new version 2.0, gpg signing 
the commit tag, and gpg signing the binary uploads, all for the first time 
after almost a year of intermittent snapshot dated releases calling them 
"year-month-day".

Here are the compressed results, divided into 1.0 (i.e. from the older 
binary-only version of Font Validator) / 2.0 (the new one), texlive 
/non-texlive part of the tests, concerning libre fonts on Fedora Linux; and 
there are the *font-list* summary which just lists the problematic font names 
in each division. Just on B/W, Size 10. (I don't have the CPU power to run 5000 
fonts through the standard test, which is B/W + Gray + sub-pixel, at Size 4 to 
72 +..., it takes about 200 x
longer).

I have separated the texlive part just because they are more platform-neutral 
and likely to be reachable via multiple different channels.

If you recognize the name of any of the fonts and know the owner, please direct 
them to the flaw detail files to look at the listed flaw details.

http://htl10.users.sourceforge.net/tmp/FontVal-test-results-2016July/

The older binary-only Font Validator 1.0 flags about 1/3 of libre fonts on 
fedora linux as problematic; the new 2.0, about 15%. 

The 2.0 result is just fewer - and possibly less annoying to some, not to chase 
phantom flaws which aren't there. Two fonts are in the 2.0 list which were not 
in the 1.0 - mona-VLGothic.ttf and wqy-microhei.ttc . The former get flagged 
because it is a bitmap font being tested at size 10 which it does not provide, 
so please ignore. The latter I do not know why - most of the glyph-level flaws 
appear to be genuine. I'd be a bit annoyed, for example, if the font developers 
put in hinting instructions to detect what host platform the font is being used 
on, to pass in 1.0 . Anyway, the font reports between 1.0 and 2.0 largely 
agrees, so I am happy to call the new version Font Validator 2.0, finally.

Code at[1], windows/mono binaries at [2] and the donate button at [3] , as 
usual, and please do click and make a donation if you find it useful. I still 
haven't got round to making the Mac OS X binaries, despite saying so last week 
- tell you a little secret: the windows binary runs about 2 to 3 times faster 
in wine + dotnet on linux [4], compared to running with native mono. And plus 
the inconvenience of a patched non-system freetype, it just seems hardly worth 
the effort to go native, merely to run slower - but I'll make the Mac OS X 
binaries one day. oh, BTW, Font Validator 2.0 is about 9 times faster than 1.0, 
on windows. I do not know why, but really glad.

[1] https://github.com/HinTak/Font-Validator
[2] 
https://sourceforge.net/projects/hp-pxl-jetready/files/Microsoft%20Font%20Validator/
[3] https://sourceforge.net/p/hp-pxl-jetready/donate/

[4] Note the default wine configuration is wine + wine-mono, not wine + dotnet .