Re: Centralize use of PG_INTXX_MIN/MAX for integer limits

>>>>> "Michael" == Michael Paquier <michael(at)paquier(dot)xyz> writes:

Michael> Hi all,

Michael> A couple of years ago, 62e2a8dc has introduced in c.h a set of
Michael> limits (to fix some portability issues from 83ff1618) to make
Michael> the code more system-independent. Those are for example
Michael> PG_INT32_MIN, etc. The core code now mixes the internal PG_
Michael> limits with the system ones. Would we want to unify a bit the
Michael> whole thing and replace all the SHRT_MIN/MAX, LONG_MIN/MAX and
Michael> such with the internal limit definitions?

Of course not. And LONG_MIN/MAX is the obvious example of why not, since
that one does vary between platforms.

INT_MAX is for the max value of an "int". PG_INT32_MAX is for the max
value of an "int32". PG_INT64_MAX is for the max value of an "int64".
LONG_MAX is for the max value of a "long". Simple.

However, as I said at the time of those patches, I did not at that stage
audit all the uses of INT_MIN/MAX to determine which should really have
been INT32_MIN/MAX. Currently, all sorts of things will likely break if
"int" and "int32" are not exactly the same size, but it might still be a
good idea to fix that at some point. (As a recent example,
contrib/intarray uses "int" almost universally, regardless of the fact
that it's meant to be dealing with SQL "integer" values, which should be
int32.)

--
Andrew (irc:RhodiumToad)