On Fri, 2005-07-15 at 14:00 +0200, Cor Bosman wrote:
. UID search ALL FROM "phyllis" .. . UID SEARCH CHARSET "US-ASCII" ALL FROM "phyllis"
Since the CHARSET applies to the search criteria, and not to the header that is searched, isnt this supposed to return the same UIDs?
Yes. Dovecot's internal handling is different though. If you don't specify character set, it compares the strings directly. If you do, it attempts to do conversions to UTF-8 for both search key and each existing header.
If you didn't compile with iconv support, or if it didn't support windows-1251 charset I guess this could have happened. I'll change the code so that it tries to compare it even if the charset translation fails, so ASCII comparing at least works correctly. See if it works in upcoming test79 or CVS.