merge mark positions incorrect

[miloush]

Merging Noto Sans Tamil Supplement, Noto Sans Tangsa and Noto Sans results in some of the combining diacritics broken:

Expected vs Actual:

It does not reproduce when merging one or other with Noto Sans, it has to be all three.
It does not reproduce when Tangsa is first and Tamil Supplement is second.

pyftmerge NotoSansTamilSupplement\googlefonts\ttf\NotoSansTamilSupplement-Regular.ttf  NotoSansTangsa\googlefonts\ttf\NotoSansTangsa-Regular.ttf NotoSans\googlefonts\ttf\NotoSans-Regular.ttf

Rendered string: r̥l̥ (r and l with ring below)

[behdad]

Unfortunately the merger is unmaintained and very limited. I don't think I'll have the time to look into this :(.

[miloush]

No worries. Can you tell me a bit about OTLOffsetOverflowError, how is it being fixed?

[behdad]

Can you tell me a bit about OTLOffsetOverflowError, how is it being fixed?

That is automatically taken care of by the compiler, by using Extension lookups, or splitting subtables into smaller ones. That shouldn't be your problem if the merge succeeds.

[behdad]

There's some mark-attachment-related issues with merger. I think you might be hitting that.

#3407

[miloush]

My obervations so far: The Tangsa And Tamil Supplement have the same anchor points:

The Noto Sans has two extra:

It is unclear to me why the order of the first two matters yet. The merged font has only the points in the first image. The bottom attachment point is first in all cases.

[miloush]

To clarify, all 3 fonts have a glyph for r. Given that I tried to merge two fonts with identical glyphs 0-9 only and they were both included in the new font, i.e. not deduplicated, I suspect the mark issue is a consequence of this - the features simply not applying to the glyphs that won cmap.

[behdad]

Sounds like it.