Order of manual mkmk and automatic mark

[simoncozens]

This is similar to notofonts/latin-greek-cyrillic#177 but different. Noto Loopless Thai has a manual mkmk feature but a "fully automatic" (i.e. not in the Glyphs source at all) mark feature. ufo2ft generates the mark code and, because there is no insert point, places it at the end of the feature file. It therefore gets compiled and processed after the mkmk code, and everything is broken.

Since there is (except perhaps in some truly pathological cases) a hard dependency on mkmk being processed after mark, we should do the right thing in the default case and ensure that automatic mark code gets inserted before any mkmk code in the feature file.

[anthrotype]

yeah, I guess it'd make more sense to insert mark before this hand-written mkmk instead of appending at the end when no explicit insertion marker is present.

[simoncozens]

I think that notofonts/latin-greek-cyrillic#274 only partially fixed this bug. I'm still seeing problems with a manual mkmk and fully automatic mark. Investigating.

[khaledhosny]

FWIW, I fix these ordering issues by having an empty mark (or mkmk) feature that only contains the insert automatic code comment.

[simoncozens]

Yeah, I do that too but it's a workaround.

[anthrotype]

I think that #512 only partially fixed this bug... Investigating

@simoncozens is this still an issue?

[simoncozens]

Yes, it's causing notofonts/lepcha#1

[simoncozens]

...and when I tried to fix it by adding an empty mark feature, I got

fontmake: Error: In 'src/NotoSansLepcha.glyphs' -> 'master_ufo/NotoSansLepcha-Regular.designspace' -> 'master_ufo/NotoSansLepcha-Regular.ufo': Compiling UFO failed: <features>:95:35: Unknown markClass @MC_Anchor10

This is because Lepcha uses abvm/blwm, and ufo2ft had generated the following erroneous code:

feature abvm {
    lookup abvm_mark2base {
        pos base uni1C23
            <anchor 305 583> mark @MC_Anchor10;
...
} abvm;

markClass uni1C2D_1C36 <anchor 0 583> @MC_Anchor10;
...
feature mark {
    lookup mark2base {
        pos base uni25CC
            <anchor 253 583> mark @MC_Anchor10
            <anchor 253 0> mark @MC_Anchor12;
    } mark2base;

} mark;

[anthrotype]

I'm closing tabs -- is this still open?

[simoncozens]

Yep.

[anthrotype]

Yes, it's causing googlefonts/noto-fonts#2321

@simoncozens mind updating this url, it points nowhere -- thanks

[simoncozens]

This is now causing Noto Sans Devanagari (plus Latin subset) to fail to build:

feature abvm {
# Automatic Code Start

} abvm;

feature blwm {
# Automatic Code Start

} blwm;

is turning into:

} vatu;

feature abvm {
    lookup abvm_mark2base {
        pos base udeva
            <anchor 299 0> mark @MC_Anchor8;
...

markClass nuktadeva <anchor -283 0> @MC_Anchor8;
...

[simoncozens]

Swapping the order of abvm/blwm in the feature file somehow managed to make things worse:

feature blwm {
    # Automatic Code Start
} blwm;

feature abvm {
    lookup abvm_mark2base {

...
} abvm;

markClass binducandradeva <anchor -130 622> @MC_Anchor2;
markClass candrabindudeva <anchor -130 622> @MC_Anchor2;
markClass anusvaradeva <anchor -126 622> @MC_Anchor2;