Anchor propagation for ligatures is broken somtimes

[khaledhosny]

Consider the font JosefinSans.glyphs; the glyph f_ij has two components f and ij and it ends up with the anchors bottom, bottom_1, bottom_2, ogonek, top, top_1, and top_2, but the top and bottom anchors make no sense here AFAIK. Even worse, the glyph f_f_ij ends with different set of anchors (name names but different positions) on each run.

Furthermore, I think the number of ligature anchors need to reflect the number of characters that make up the ligature not the number of components which may or may not be unrelated. i.e. if we have sub f ij by f_ij; then there should be two ligature anchors regardless of how many components f_ij glyph has.

[behdad]

Furthermore, I think the number of ligature anchors need to reflect the number of characters that make up the ligature not the number of components which may or may not be unrelated. i.e. if we have sub f ij by f_ij; then there should be two ligature anchors regardless of how many components f_ij glyph has.

Actually the way OpenType is implemented in Uniscribe and HarfBuzz, if ij was itself a ligature of two components, the f_ij will get three components.

[khaledhosny]

I stand corrected then.

[moyogo]

Well, f_ij is, according to AGLFN, the ligature between the two characters f and ij U+0133. That is also what the font does in the dlig sub f ij by f_ij;.

[anthrotype]

looks like a duplicate of #368