Brand-new in PG 7.2 is "create or replace function". This has the advantage of not breaking other compiled functions which refer to your function. Oracle's had this forever (for the same reason).
When we finally ditch PG 7.1 support (later this year, not in the near term, of course!) we'll get someone to write a little sed script to replace all of the "create function" statements with "create and replace".
Anyway, yes, the ::char was a mistake, uncaught by PG 7.1 and earlier. In reality it was treating it like a varchar but it was still an error.
If someone wants to gather up a patch and put it into the SDM it would be deeply appreciated ...
