On Tue, Jun 14, 2022 at 01:23:04PM -0400, Tom Lane wrote:
> I suppose it would be too user-unfriendly to generate those scripts
> underneath $PGDATA, so don't we need to put back the .gitignore
> entries?
delete_old_cluster.sh is generated within tmp_check/ in a non-VPATH
build:
$ cd src/bin/pg_upgrade/ && make check
$ find . -name delete_old_cluster.sh
./tmp_check/delete_old_cluster.sh
$ git grep chdir -- *.pl
t/002_pg_upgrade.pl:chdir ${PostgreSQL::Test::Utils::tmp_check};
Or you have a workflow where you do a "make clean" after running the
tests on a past stable branch? This would leave behind
delete_old_cluster.sh, especially if the cleanup is triggered on
HEAD. With each branch taken in isolation, there is no need for a
.gitignore entry. (FWIW, I just use a worktree these days in my flow
of patching across multiple stable branches.)
--
Michael