Hello,
while upgrading a single journal installation I was able to convert latin1 characters stored in utf8 database to utf8 with exception of submission_search_keyword_text table. Apparently it’s impossible to convert it to binary and then convert to utf8 with utf8_unicode_ci collation as mysql won’t allow for duplicates in this table (and there happen to be duplicates with this collation since original data was stored as latin1 in utf8 table). I tried to circumvent it by setting collation for this table to utf8_bin but then the upgrade script will throw an error (duplicate detected) and quit.
What is submission_search_keyword_text actually for? I can see a ton of keywords but it seems that those aren’t keywords actually stored with articles as those can be found in different tables.
It looks like a sort of index for search purposes. If so, can it be regenerated after upgrade to OJS3?