I am going to withdraw this proposal. In another email thread, we
learned two very important things:
1. The problem of a locale names changing for an *existing* database
can apparently be fixed using "Locale Builder"[1]. Users in Turkey
have presumably already addressed the situation that way and there are
reports of articles about that in Turkish. We don't really need the
"setlocale.map" rename-on-the-fly patch in this thread, and I don't
want to maintain it if we don't need it.
2. The Windows runtime library is crashing when initdb's subprocess
postgres.exe calls it with a name encoded in an encoding it doesn't
like[2]. (IMHO it should return NULL instead! The C standard does
not have a demons-may-fly-out-of-your-nose clause.) We need to tell
users not to use those locale names. That string is actually coming
from the EDB installer, and we'll follow up on that. Perhaps
PostgreSQL should also consider stripping non-ASCII characters to
force it to fail nicely.
[1] https://www.postgresql.org/message-id/CA%2BhUKGLyfou2xa2fxGOs0WwkOrcKjwqn-bRNR01Hq5RAD16ueg%40mail.gmail.com
[2] https://www.postgresql.org/message-id/CA%2BhUKG%2BFxeRLURZ%3Dn8NPyLwgjFds_SqU_cQvE40ks6RQKUGbGg%40mail.gmail.com
