Wasn't this patch
http://archives.postgresql.org/pgsql-committers/2007-11/msg00170.php
supposed to fix things so that we wouldn't be throwing NOTICEs out of
tsearch dictionary init functions? It seems to have gotten only
one of the two elog(NOTICE) calls in there. The other one is for a
stopword in the replacement phrase. Is there any objection to making
that an ERROR too?
Also, several of the elog's in this file ought to be ereport's since
they are user-caused error conditions. I think we are nominally past
string freeze, but does anyone have an objection to changing them to
ereport's? If they don't get translated, the worst that will happen
is that they come out in English, which is what's happening now.
regards, tom lane