Re: Cutting test runtime for src/test/modules/snapshot_too_old

On Tue, Aug 2, 2022 at 11:38 AM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> I've complained before that the snapshot_too_old TAP tests seem
> ridiculously slow --- close to a minute of runtime even on very fast
> machines. Today I happened to look closer and realized that there's
> an absolutely trivial way to cut that. The core of the slow runtime
> is that there's a "pg_sleep(6)" in the test case; which perhaps could
> be trimmed, but I'm not on about that right now. What I'm on about
> is that two of the three isolation tests allow the isolationtester to
> default to running every possible permutation of steps, one of which
> doesn't even generate the "snapshot too old" failure. IMV it's
> sufficient to run just one permutation. That opinion was shared by
> whoever wrote sto_using_hash_index.spec, but they didn't propagate
> the idea into the other two tests.
>
> The attached cuts the test runtime (exclusive of setup) from
> approximately 30+24+6 seconds to 6+6+6 seconds, and I don't see
> that it loses us one iota of coverage.
>
> I cleaned up some unused tables and bad comment grammar, too.

Yeah, I feel like it was a mistake to allow the list of permutations
to be unspecified. It encourages people to just run them all, which is
almost never a thoughtful decision. Maybe there's something to be said
for running these tests in one successful permutation and one failing
permutation -- or maybe even that is overkill -- but running them all
seems like a poor idea.

--
Robert Haas
EDB: http://www.enterprisedb.com