> The odd thing is that the first five ligatures -- the ones > groff is supposed to be able to recognize when written in > plain text -- are the ones that look the same in both their > non-ligature and ligature forms, while the others are visibly > different. So I wonder if groff itself is somehow thwarting > the ligature process -- it seems a particularly interesting > coincidence that only the five groff-approved ligatures aren't > working correctly. (If I expand my test case to include > all the ligatures defined in the Libertine Roman font, it is > still only those five that don't behave as expected.)
Yes, it appears that groff decomposes the "u..." characters, recognizes the five as known ligatures, but then outputs the individual characters because the font description file doesn't contain the corresponding groff names "ff", "Fi", "Fl", etc., even though it contains the composite unicode characters. For those it doesn't recognize as one of the five defined ligatures, it outputs the unicode character as specified in the font description file. Have you tried adding "f_i", "f_f_i", etc. to the "textmap" and "text.enc" files?
