Hello Peter,
>> About the pg_dump code, I'm wondering whether it is worth generating
>> LOCALE as it breaks backward compatibility (eg dumping a new db to load it
>> into a older version).
>
> We don't really care about backward compatibility here. Moving forward,
> with ICU and such, we don't want to have to drag around old syntax forever.
We will drag it anyway because LOCALE is just a shortcut for the other two
LC_* when they have the same value.
> How about this patch?
It applies cleanly, compiles, global & pg_dump make check ok, doc gen ok.
I'm still unconvinced of the interest of breaking backward compatibility,
but this is no big deal.
I do not like much calling strlen() to check whether a string is empty,
but this is pre-existing.
I switched the patch to READY.
--
Fabien.