Tests "with" and "alter_table" suffer from name clash

Thomas Munro <thomas.munro@gmail.com> writes:

With unlucky scheduling you can get a failure like this:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hoverfly&amp;dt=2021-12-22%2010%3A51%3A32
Suggested fix attached.

Looks reasonable. We really should avoid using such common
names for short-lived tables in any case --- it's an invitation
to trouble. So I'd vote for changing the other use of "test", too.

regards, tom lane

On Thu, Dec 30, 2021 at 3:27 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Looks reasonable. We really should avoid using such common
names for short-lived tables in any case --- it's an invitation
to trouble. So I'd vote for changing the other use of "test", too.

In fact only REL_10_STABLE had the problem, because commit 2cf8c7aa
already fixed the other instance in later branches. I'd entirely
forgotten that earlier discussion, which apparently didn't quite go
far enough. So I only needed to push the with.sql change. Done.

Thomas Munro <thomas.munro@gmail.com> writes:

In fact only REL_10_STABLE had the problem, because commit 2cf8c7aa
already fixed the other instance in later branches. I'd entirely
forgotten that earlier discussion, which apparently didn't quite go
far enough. So I only needed to push the with.sql change. Done.

Hah, I thought this felt familiar. So the real problem is that
my backpatch (b15a8c963) only fixed half of the hazard. Sigh.

regards, tom lane