inoue(at)postgresql(dot)org writes:
> Modified files:
> src/include/parser: parse_coerce.h
> src/include/catalog: pg_operator.h
> Log message:
> Allow comparison between xid and xid, int.
I do not think it's a good idea to define xid and int as
binary-compatible. If it is a good idea then you did it wrong
--- IS_BINARY_COMPATIBLE needs to test for both directions.
But what I'd suggest is to undo the IS_BINARY_COMPATIBLE change
and instead add another pg_operator entry that accepts xid = int4,
mapping it to xideq. That gets the job done without opening up
a morass of binary-compatibility issues.
regards, tom lane