While reviewing another patch, I was playing around with a test case
to trigger a large memory allocation. I was doing this on Windows when
I got a nasty looking WARNING in a MEMORY_CONTEXT_CHECKING build:
create table t (a int);
insert into t values(1);
alter table t alter column a set (n_distinct = -1); -- all values distinct
analyze t;
update pg_class set reltuples = 1e9 where oid = 't'::regclass; -- hack
to make the table big
set work_mem = '4GB';
explain (summary on) select a from t except select a from t;
and got:
WARNING: problem in alloc set ExecutorState: bad single-chunk
0000023DE7C98098 in block 0000023DE7C98070
WARNING: problem in alloc set ExecutorState: bad single-chunk
0000023DE7C98098 in block 0000023DE7C98070
It turns out that AllocSetCheck() thinks "long" is a good datatype to
store the difference between 2 pointers. That's not going to work well
on 64-bit Windows as long is 32-bit.
I did also consider [u]intptr_t, but thought Size was better as that's
what chsize is.
Trivial fix attached.
David
