Re: shadow variables - pg15 edition

On 2022-Oct-06, David Rowley wrote:

> On Thu, 6 Oct 2022 at 02:34, Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org> wrote:
> > A simpler idea might be to just remove the inner declaration, and have
> > that block set the outer var. There's no damage, since the block is
> > going to end and not access the previous value anymore.
> >
> > diff --git a/src/bin/pgbench/pgbench.c b/src/bin/pgbench/pgbench.c
> > index aa1a3541fe..91a067859b 100644
> > --- a/src/bin/pgbench/pgbench.c
> > +++ b/src/bin/pgbench/pgbench.c
> > @@ -7506,7 +7506,7 @@ threadRun(void *arg)
> > /* progress report is made by thread 0 for all threads */
> > if (progress && thread->tid == 0)
> > {
> > - pg_time_usec_t now = pg_time_now();
> > + now = pg_time_now(); /* not lazy; clobbers outer value */
>
> I didn't want to do it that way because all this code is in a while
> loop and the outer "now" will be reused after it's set by the code
> above. It's not really immediately obvious to me what repercussions
> that would have, but it didn't seem worth taking any risks.

No, it's re-initialized to zero every time through the loop, so setting
it to something else at the bottom doesn't have any further effect.

If it were *not* reinitialized every time through the loop, then what
would happen is that every iteration in the loop (and each operation
within) would see exactly the same value of "now", because it's only set
"lazily" (meaning, if already set, don't change it.)

--
Álvaro Herrera Breisgau, Deutschland — https://www.EnterpriseDB.com/