On Wed, Jun 22, 2022 at 11:03:22AM -0400, Andrew Dunstan wrote:
> On 2022-06-22 We 03:21, Noah Misch wrote:
> > On Tue, Apr 19, 2022 at 07:24:58PM -0400, Andrew Dunstan wrote:
> >> On 2022-04-19 Tu 18:39, Michael Paquier wrote:
> >>> +*generate_ascii_string = *TestLib::generate_ascii_string;
> >>> +*slurp_dir = *TestLib::slurp_dir;
> >>> +*slurp_file = *TestLib::slurp_file;
> >>>
> >>> I am not sure if it is possible and my perl-fu is limited in this
> >>> area, but could a failure be enforced when loading this path if a new
> >>> routine added in TestLib.pm is forgotten in this list?
> >> Not very easily that I'm aware of, but maybe some superior perl wizard
> >> will know better.
> > One can alias the symbol table, like https://metacpan.org/pod/Package::Alias
> > does. I'm attaching what I plan to use. Today, check-world fails after
> >
> > sed -i 's/TestLib/PostgreSQL::Test::Utils/g; s/PostgresNode/PostgreSQL::Test::Cluster/g' **/*.pl
> >
> > on REL_14_STABLE, because today's alias list is incomplete. With this change,
> > the same check-world passes.
The patch wasn't sufficient to make that experiment pass for REL_10_STABLE,
where 017_shm.pl uses the %params argument of get_new_node(). The problem
call stack had PostgreSQL::Test::Cluster->get_new_code calling
PostgreSQL::Test::Cluster->new, which needs v14- semantics. Here's a fixed
version, just changing the new() hack.
I suspect v1 also misbehaved for non-core tests that subclass PostgresNode
(via the approach from commit 54dacc7) or PostgreSQL::Test::Cluster. I expect
this version will work with subclasses written for v14- and with subclasses
written for v15+. I didn't actually write dummy subclasses to test, and the
relevant permutations are numerous (e.g. whether or not the subclass overrides
new(), whether or not the subclass overrides get_new_node()).
> Nice. 30 years of writing perl and I'm still learning of nifty features.
Thanks for reviewing.