Since one point in Glyphs 2, using Adobe-Identity-0 ROS started to NOT assign cid+1 unless you specify in glyphOrder custom parameter starting with “.notdef, space…”. Is there a purpose in this change or is it simply a bug?
What I understand is that when ROS-0 is set, the font is almost like a name-keyed (not double encoding glyphs, one dict array) with the exception that the cid will be assigned to the gid if the character doesn’t have a unicode that is related to a real cid.
e.g. Figure out that you want to create a TTC sharing the same CFF, for that you will need to create a Glyphs file with all the glyphs that is going to be use for all fonts of the same weight, in order to recycle the kanjis.
So, if in one font you have all glyphs to be use in a “regular” font, but then you add at the end of the CFF (after all glyphs for one font) only the kanas to act as the default for another font specially designed for a small size, which only is going to have different drawings in kanas but it is going to preserve all other glyphs. So you will prepare two instances, one the regular with the regular encoding, and one instance to generate the special optical variation, in which you need to re-encode all kanas in order to assign the code point to those kanas that you have added at the end. In theory you will have two fonts that are almost identical (you should change the style name) except for the cmap and name table.
Ok what I have noticed is that for those additional non-encoded kanas that are going to be used in the optical variation, in the regular font were this glyphs are unencoded, the cid = gid, but in the font that is generated with this kanas encoded, there are not cid in the font.ps, in fact the font.ps is exported using the glyph name.
So @Tosche, sometime it is working as expected and sometimes not. My previous example is a test case to see the is a isn’t working the cid assignment in ROS-0