defcon.errors.DefconError: The kerning data is not valid while processing NotoSansEthiopic-MM.glyphs about fontmake HOT 14 CLOSED

Same error happens when:
==== building src/NotoSerif-ItalicMM.glyphs ====
==== building src/NotoSerif-RomanMM.glyphs ====

Inputs also for these two additional builds are attached.
NotoSerif.zip

from fontmake.

I think the problem here is that the Glyphs source contains conflicting kerning rules, which are rejected on UFO load by defcon. We didn't have this problem with RoboFab (the conflicting kerning rules are removed when the feature is written), but interpolation via MutatorMath uses defcon so we have to deal with this earlier than on feature write.

@adrientetar do you know of some tool that can remove conflicting kerning in defcon or elsewhere?

from fontmake.

I don't.

Btw kerning is normally loaded lazily but needs to be processed on open when converting from a legacy ufo version.

from fontmake.

Hm. This is still a bit of a pain to deal with because robofab doesn't support ufo3.

FYI @behdad it's starting to look like we need to scrap use of robofab universally in favor of defcon, if only for compatibility with MutatorMath (this would also apply to Roboto, if we want to use it there).

from fontmake.

Sounds like more work than we previously planned. However, I'm all for
doing "the right thing".

On Wed, Feb 17, 2016 at 3:06 PM, James Godfrey-Kittle <
[email protected]> wrote:

Hm. This is still a bit of a pain to deal with because robofab doesn't
support ufo3.

FYI @behdad https://github.com/behdad it's starting to look like we
need to scrap use of robofab universally in favor of defcon, if only for
compatibility with MutatorMath (this would also apply to Roboto, if we want
to use it there).

—
Reply to this email directly or view it on GitHub
#13 (comment).

Marek Z Jeziorek [ 老马 ] | [email protected] | 312 725-6958

from fontmake.

Great idea (cc @graphicore), ufo3 has things like layers standardized so you don't need to use robofont extensions to the spec anywhere. defcon itself has many refinements in the ufo3 branch and, as you found out, validates data.
If you're doing this I'll work on getting our patches upstreamed.

from fontmake.

👍 for dropping robofab in favour of defcon+ufo3!

from fontmake.

FYI @behdad it's starting to look like we need to scrap use of robofab universally in favor of defcon, if only for compatibility with MutatorMath (this would also apply to Roboto, if we want to use it there).

Correct.

I'll take a look into this issue. I'm new to defcon, but see what I can figure out.

from fontmake.

But, kerning is still validated by defcon on UFO save so using defcon doesn't really help us here. It's looking to me like we have no choice but to remove the duplicate kerning earlier if we use MutatorMath. Of course we can still switch to defcon since it seems like the right thing to do.

from fontmake.

So, is the kerning data in one of the masters invalid? Then that's a source bug. @jamesgk can you make defcon report the invalid pairs? validateKerning definitely generates those, but they are lost in the exception.

It also occurs to me that:

1. defcon's validateKerning can be improved a bit. I'll send PR for that separately,
2. mathFont.mathKerning._processMathOne() needs to be more sophisticated to handle adding two kernsets with differing groups, half-class exceptions, etc. I'll send a PR for that as well as I work it out this week.

from fontmake.

So, is the kerning data in one of the masters invalid? Then that's a source bug.

So, is it OK if fontmake chokes on invalid kerning? My concern is that Glyphs does not.

@jamesgk can you make defcon report the invalid pairs? validateKerning definitely generates those, but they are lost in the exception.

They are passed on by defcon via an attribute of the exception object; we can print that out in glyphs2ufo where mutatorMath is called.

from fontmake.

BTW, there are quite a lot (948) of these conflicting rules in Noto Sans Ethiopic. The first 10:

public.kern1.@MMK_L_li_ethiopic, qa_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qu_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qi_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qee_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qe_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qo_ethiopic (-60) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qwa_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qwi_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qwe_ethiopic (-70) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qho_ethiopic (-73) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
...

from fontmake.

Thanks. I'll take a look.

On Tue, Mar 1, 2016 at 8:27 AM, James Godfrey-Kittle <
[email protected]> wrote:

BTW, there are quite a lot (948) of these conflicting rules in Noto Sans
Ethiopic. The first 10:

public.kern1.@MMK_L_li_ethiopic, qa_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qu_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qi_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qee_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qe_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qo_ethiopic (-60) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qwa_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qwi_ethiopic (-140) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qwe_ethiopic (-70) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
public.kern1.@MMK_L_li_ethiopic, qho_ethiopic (-73) conflicts with hhi_ethiopic, public.kern2.@MMK_R_qa_ethiopic (-62)
...

—
Reply to this email directly or view it on GitHub
#13 (comment).

behdad
http://behdad.org/

from fontmake.

The original issue here should be fixed. New issues can be spun off e.g. for migration to defcon (one already exists in ufoLib for updating validateKerning).

from fontmake.