fontforge 220308 test fail

[kilobyte]

As reported in Debian:

CHECKING build/3270-Regular.otf
CHECKING 98 Self-intersecting glyph (issue #2) when FontForge is able to
correct this (overrides general setting)
ERROR 2 Self-intersecting glyph
ERROR 5 Missing points at extrema
FAIL build/3270-Regular.otf

[rbanffy]

The export to OTF has been creating some low-quality fonts. Would it make sense to drop support to it and only package the TTF version instead?

[kilobyte]

It seems to me that neither of the errors would be fixed by switching to TTF. I didn't test, though (not at home atm).

[rbanffy]

I'll check it out.

…

On Sat, Jun 25, 2022 at 6:33 PM Adam Borowski ***@***.***> wrote: It seems to me that neither of the errors would be fixed by switching to TTF. I didn't test, though (not at home atm). — Reply to this email directly, view it on GitHub <#110 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABNDCVAL6IXSCEDPU4F2HLVQ47H5ANCNFSM5ZXU3AIA> . You are receiving this because you commented.Message ID: ***@***.***>

-- Ricardo Bánffy http://about.me/rbanffy

[rbanffy]

Thanks, Adam. The logs on Sid show (/me slaps forehead) a newer version of Fontlint (and, presumably, FontForge) It looks like the line: for f in build/*f; do fontlint -i 98 $f || exit 1; done is what blocks the build. It may be an issue with the font itself, or with the newer Fontforge and its pieces of it that are used in fontlint. In any case, a `-i 2 -i 5` added to the line should remove the block in the build. What other tests is the Debian build running? It'd be worthwhile to port them back here (I'm looking at the Makefile and will add a for loop instead of the repeated lines I have there now) In the meantime, I'll investigate it using the newest Fontforge.

…

On Sat, Jun 25, 2022 at 6:33 PM Adam Borowski ***@***.***> wrote: It seems to me that neither of the errors would be fixed by switching to TTF. I didn't test, though (not at home atm). — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

-- Ricardo Bánffy http://about.me/rbanffy

[kilobyte]

I run only that test.

In the past, I called the upstream (yours) tests, but there's too much dependency on external network access, thus I gave up and hard-coded the above line.

[rbanffy]

I think we'll need a better font linter. In your case, I'd suggest adding -w or -i to the failing tests (which seems to be more or less harmless - I could use the fonts even with the issues pointed out by fontlint). I am also going for a new release of the font, as I figured out a nice way to make the condensed versions actually condensed rather than just clip the character cell. There will be some longer term fine tuning but the improvement is already quite visible, in particular for the block graphics.

…

On Sun, Jun 26, 2022 at 3:31 PM Adam Borowski ***@***.***> wrote: I run only that test. In the past, I called the upstream (yours) tests, but there's too much dependency on external network access, thus I gave up and hard-coded the above line. — Reply to this email directly, view it on GitHub <#110 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABNDCTUKZBRYYVA4C73XZ3VRBSVLANCNFSM5ZXU3AIA> . You are receiving this because you commented.Message ID: ***@***.***>

-- Ricardo Bánffy http://about.me/rbanffy

[kilobyte]

I've added -i for now, thanks for the explanations.

[davide125]

While updating this in Fedora to 3.0.0 I hit the same issue described here. After adding -i 2 -i 5 I have one last error:

+ fontlint -i 98 -i 2 -i 5 build/3270SemiCondensed-Regular.otf
Copyright (c) 2000-2022. See AUTHORS for Contributors.
 License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
 with many parts BSD <http://fontforge.org/license.html>. Please read LICENSE.
 Version: 20220308
 Based on sources from 2022-06-13 00:00 UTC-ML-D-GDK3.
Attempt to unget two characters
CHECKING     build/3270SemiCondensed-Regular.otf
ERROR      3 Wrong direction
FAIL         build/3270SemiCondensed-Regular.otf

which I've worked around for now by adding -i 3.

[kilobyte]

-i 3 ("Wrong direction") wasn't enough, there's also -i 34 ("Bad CFF table"), that's why I disabled all tests altogether in the Debian packaging.

[rbanffy]

3 seems correctable in the build script. 34 seems more enigmatic (and why in hell would exporting the OTF corrupt a metadata table?!)