Greg Stark <gsstark@mit.edu> writes:
> This is presumably because of the long-standing issue that Postgres takes the
> snapshot as soon as the BEGIN is issued.
No, we don't set the snapshot until the first DML query is issued. This
is critical for serializable transactions: you have to be able to take
locks before the snapshot is frozen.
There are at least three interesting events involved: 1 BEGIN command issued 2 First lock taken (presumably as a
consequenceof another command) 3 Transaction snapshot taken (ditto; might be a different command)
We have to start the transaction no later than event #2 since there has
to be something to hold the lock. But it'd be easy enough to decouple
this from BEGIN, and it'd be good enough to solve the "COMMIT;BEGIN"
problem.
Which of these three times do you think now() ought to correspond to?
I recall having argued that it should be event #3 since that corresponds
to the database snapshot you see. 100% backwards compatibility would
require setting now() at event #1, but will anyone weep if we change that?
regards, tom lane