In commit http://cgit.freedesktop.org/libreoffice/core/commit/?id=6e225b41f1ab3e6cac395b0c0c6db73414658625 and bug https://bugs.documentfoundation.org/show_bug.cgi?id=55707 changes were made to Finnish word counting by removing previous customisations to breakiterator rules. Unfortunately while these changes improved word counting they had negative effect on spell checking: we should use same word boundaries for spell checking and word counting but this is no longer the case. For example the test case string "Kuorma-auto kaakkois- ja Keski-Suomi USA:n 90:n %:n" should count as 7 words for word counting (as it correctly does) but these same 7 words should also be used for spell checking. So the words for spell checking should be Kuorma-auto (works correctly) kaakkois- (now incorrectly just kaakkois) ja (works correctly) Keski-Suomi (works correctly) USA:n (now incorrectly split at colon) 90:n (now incorrectly split at colon) %:n (now incorrectly split at colon) I have filed a change request to CLDR to have the hyphen error fixed there: http://unicode.org/cldr/trac/ticket/8368
Caolan: thought you might be interested in this one.
I'd consider this more an "closed->upstream" sort of thing as we can't sustainably support customized icu rules, its too hard as icu/unicode evolves over time to keep them right.
If it's mainstream, let's put this one to notourbug. Perhaps it'll be fixed in 5.5.1 see http://site.icu-project.org/ (since we use ICU 5.4.1) Anyway, thank you Caolan for your feedback.