On Tue, Sep 24, 2024 at 10:28 AM Tom Lane <tgl@xxxxxxxxxxxxx> wrote:
It's even looser than that, really: it's the first statement that
requires an MVCC snapshot.
Hm....so why does "SELECT 1;" work as a transaction start marker then, as opposed to "SHOW work_mem;", which does not? Do we simply consider anything with a SELECT as needing a snapshot?
Cheers,
Greg