On 8/15/25 8:00 PM, Philipp Kern wrote:
I narrowed it down to another row in hurd-i386 in sid. But I don't know what the row is. I assume some email reply made it with invalid UTF8 into the table. And it's hilariously complicated to figure out what the row is because the postgres client just explodes when it sees invalid content.`\encoding sql_ascii' in the client makes it work.If you can spot which row is at fault I'm happy to fix it. And maybe we can recover the content this time and track back what happened.
Cleared out a couple more (5). Now it works. But the database is in SQL_ASCII, so setting the client to SQL_ASCII (e.g. by setting PGCLIENTENCODING=SQL_ASCII in the environment, thanks Myon!) might make sense. It's not guaranteed to be UTF-8 at this point. :/
Kind regards Philipp Kern