On 2/17/25 17:21, Arthur Rosendahl wrote:
[...] The first example yields ἅπα-ντα (incorrect according to the patterns), the second one ἅπαν-τα (correct up to \lefthyphenmin). Hans, please comment :-) The result should be the same in both cases, since the Spanish and Ancient Greek patterns don’t interfere with each other -- and in any case the example with only the Greek patterns is already incorrect.
I think I (partially) discovered what is misbejaving here. In lang-def.{mkxl|mkiv}, ancient Greek is defined: \installlanguage [\s!agr] [\s!default=\s!gr, \c!leftquote=\upperleftsinglesixquote, \c!rightquote=\upperrightsingleninequote, \c!leftquotation=\upperleftdoublesixquote, \c!rightquotation=\upperrightdoubleninequote] If a last line loading the patterns isn’t added (\s!patterns=\s!agr), monotonic Greek patterns are loaded. With that change, \registerhyphenationpattern and \registerhyphenationexception work for \agr, but not for secondary loaded patterns (such as in \setuplanguage[en][patterns={en, agr}]). It seems that cases such as \setuplanguage[en][patterns={en, agr}], exceptions to the main patterns aren’t loaded at all. Hans, could you confirm this? Many thanks for your help, Pablo