|
From: Ralf S. <ral...@we...> - 2006-09-27 18:45:38
|
On Wed, Sep 27, 2006 at 08:45 -0700, George Williams wrote: > On Mon, 2006-09-25 at 14:38, Ralf Stubner wrote: > > Ok, some more details. I have the sources for Junicode 0.6.11 here. T= He > > sfd file has several GSUB features, first ones being 'rtla', 'smcp', > > 'c2sc', and 'liga'. LAter on there also is an incorrect 'liag' featur= e=B9 > > If I generate a TTF from these sources using ff based on sources from > > 08:15 22-Aug-2006 the resulting font has a different order of the GSU= B > > features. It looks like they have been sorted alphabetaically whhich > > means that 'smcp' comes after 'liga'. This causes problems when one > > tries to use the 'smcp' feature on a word like 'file', since first th= e > > 'fi' ligature is formed and only after that small caps are used. See > > Pablo's message from today. > As far as I can tell fontforge outputs features in the order you tell > fontforge to output those features. If you have a counter-example pleas= e > send me a test case. Right now the only example I have is Junicode version 0.6.11 from <URL:http://junicode.sf.net/>. If I open Junicode-Regular.sfd I see a feature list starting with 'rtla', 'smcp', 'c2sc', 'liga' .... Generating a TTF from this I get a feature list starting with ' RQD', 'c2sc', 'ccmp', .... This is the same ordering of features as present in the TTFs available at the above site. I have tried to build a smaller example but haven't succeeded yet. BTW, Junicode-Regular.sfd contains: TableOrder: GPOS 0 TableOrder: GSUB 0 Might this be related? How is feature order decided in such a case? cheerio ralf |