Greg Stark <gsstark@mit.edu> writes:
> Well if all we want to do is reproduce the current functionality of EXPLAIN
> ANALYZE, all you need is a single sig_atomic_t int that you store the address
> of the current node in.
Do I need to point out that sig_atomic_t can't portably be assumed to be
wider than char?
We do currently assume that TransactionId can be read/written
atomically, but (a) that's an int not a pointer, and (b) the assumption
is cruft that we really ought to get rid of.
In any case, speculating about whether we can do something useful with
atomic types ignores the main real problem the thread is about. Anybody
remember process titles and current command strings?
regards, tom lane