Sorry, I didn't notice the filename went into the $dsn -- I'm not a database expert, and didn't know it could go there. Digging into the code, it appears that it's parsed via the underlying c code, using mysql_options(sock, MYSQL_READ_DEFAULT_FILE, df);. The mysql-options documentation says that you use that option to get it to read from the given file... The option-files doc claims that "Leading and trailing spaces are automatically deleted from option names and values", which implies that spaces around the equal shouldn't matter. It may be that you've got a different version of mysql than the 8.0 reference manual my searches brought me to, or that the statement is misleading and spaces around the equal are bad.

edit: fix "optoins" typo. Also, want to clarify: when I saw the anonymous "seriously" reply, I wanted something slightly more helpful than that. However, I was apparently confused into believing the parsing was something you had done externally to what you had quoted, rather than something that mysql handles internally. So instead of being helpful, I went in the wrong debug direction. Sorry.

Hmmmmmmmmmmm............

I DID recently switch from community MySQL to MariaDB 10.2 (because it's apparently my only practical option on Gentoo Linux for a Galera cluster). Could MariaDB maybe be doing something iffy...? It wouldn't be the first behavioral change I've seen in MariaDB with no clearly apparent technical reason than to diverge from MySQL.

(That said, I do laud MariaDB for not adopting MySQL 5.6's ill-conceived and badly-implemented history-sanitization "feature".)

FWIW, there is (relatively) new DBD::MariaDB. (I do not know if it provides a configuration parser that would be more flexible to keep the file easier to read by allowing whitespace around "=" sign.)