[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[ft-devel] test result part 2 ( Re: Font Validator 2.1 release note/stat
From: |
Hin-Tak Leung |
Subject: |
[ft-devel] test result part 2 ( Re: Font Validator 2.1 release note/status, and libre font test results a year on after FontVal 2.0.) |
Date: |
Sat, 22 Jul 2017 16:47:31 +0000 (UTC) |
Forgot to mentioned this, though it is not really of any major interest to
most:
The debate or discussion around W6029 (Setting the loop variable to zero) turns
out not to be interesting
as far as the objective of "getting a pass" is concerned. Despite Werner saying
ttfautohints does that, of
the ~1400/1500 libre fonts with truetype outlines, only 21 shows that. 16 of
them have other problems
with their hinting instructions so would have failed the hinting instruction
test overall anyway;
i.e. only 5 fonts would have "passed the test" out of 1500 fonts (or 3283...).
So the issue is
unimportant either way.
The 5 fonts, as far as I can remember, are all of the Bitstream*Mono and
Vera*Mono variants, and not
really actively maintained.
There are some duplicates among the 3283 fonts as TexLive ships a fair number
of libre fonts also;
though not necessarily at the same version of the main OS.
To recap, here is the summary for the hinting instruction check:
3283 fonts
1[34]9x skipped, CFF
1[34]9x passed
394 fails to some degree
1 zero-length font file,filed at redhat's bugzilla, probably
mis-packaging.
out of the 394; 21 shows W6029; out of the 21, 5 shows ONLY W6029.
Also while testing the changes in glyf tests part of the new release (i.e. we
are not talking
about FreeType any more), I noted that the KanjiStrokeOrders font
( http://www.nihilist.org.uk/
https://fedoraproject.org/wiki/KanjiStrokeOrders_fonts )
is especially poor in engineering terms - lots of very
closely-spaced/overlapping control points,
hair pins - i.e. contours going back on itself - and some tight loops, and
other poor constructs
which causes older FontVal to crash when analysing the contours for knots and
etc. This is all fixed
in Font Val 2.1 of course, but somebody might actually like to approach the
font designer and offer
some help on making his font better...
--------------------------------------------
On Sat, 22/7/17, Hin-Tak Leung <address@hidden> wrote:
<snipped>
Below is the release notes. As I don't
have the test results with me, I can only write from
memory: out of the 3283 fonts shipped
by fedora 26 ( a slightly larger number than fedora 24
for which FontVal 2.0 was tested
against), about 139x + 149x are uninteresting CFF fonts
- which the hinting test skips, and
those that passes - I can't remember which is which, but
that adds up to 288x fonts.
394 have some degree of problems with
their hinting instructions;
this is about the same as last year, so
obviously this tool needs to be better publicised.
One was a bad packaging of an
zero-length file ( filed as
https://bugzilla.redhat.com/show_bug.cgi?id=1473437 ).
<snipped>
- Known Caveats:
...
- W6029 (Setting the loop variable to zero is an error) may be withdrawn
later as ttfauthint does it and there is no known problem, despite
some apple and macrosoft docs saying so.
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [ft-devel] test result part 2 ( Re: Font Validator 2.1 release note/status, and libre font test results a year on after FontVal 2.0.),
Hin-Tak Leung <=