Jean-Marc Lasgouttes wrote:
This trivial fix is going in now: we have to use nb_NO instead of no_NO, as we already to in po/.
I now get correct spellchecking - for words which happens to be ascii-only. This patch is a success.
I guess this is a well-known unicode problem, but anyway: Both encodings are bad: LANG=nb_NO.UTF-8 LANG=nb_NO.ISO8859-1 Spellchecking a (correct) non-ascii word makes the spellchecker suggest a replacement where the non-ascii letters are replaced by garbage. Probably the same characters in a different encoding. This results in invalid unicode showing up "strange" on screen. Lyx then _crash_ upon saving, leaving the .lyx file truncated before the bad word. I guess spellchecking simply isn't finished. I also hope the crashing will be resolved - there are many ways of getting bad unicode into a paragraph - it'd be nice if lyx didn't die as a result. Just eating the bad characters, leaving them there, or replacing with an error symbol is all better than crash+truncate. Do you want a bugzilla entry, or is this well-known unicode trouble? Helge Hafting
