Philip Warner <pjw@rhyme.com.au> writes:
>> Philip, I see you applied some pg_dump patches yesterday. Have you
>> resolved all your outstanding issues, or is there still more you want
>> to do before 7.1.2?
> Everything I know about is resolved.
Okay, good. I did some experimentation this afternoon with dumping the
7.0 regression database using both native 7.0 pg_dump and the
current-sources one. Seemed to work pretty well, though I did make one
change: I think we should assume proisstrict = FALSE when dumping from a
7.0 db, not TRUE. The forcing consideration for this is that I was
getting "isstrict" markers on plpgsql_call_handler, which would be a
really nasty problem if it got into the field: people would report that
they couldn't get plpgsql functions to work with NULLs, and we'd be
unable to duplicate the misbehavior. More generally, it doesn't matter
for old C functions, because the fmgr_oldstyle wrapper will cause the
right things to happen, and I don't think we want to force strictness
for SQL or PL functions. (That's why I chose CREATE FUNCTION's default
to be non strict...)
regards, tom lane
