pgsql: Move rolenames test out of the core regression tests. - Mailing list pgsql-committers

From Tom Lane
Subject pgsql: Move rolenames test out of the core regression tests.
Date
Msg-id E1hhd36-0000bR-Dy@gemulon.postgresql.org
Whole thread Raw
List pgsql-committers
Move rolenames test out of the core regression tests.

This test script is unsafe to run in "make installcheck" mode for
(at least) two reasons: it creates and destroys some role names
that don't follow the "regress_xxx" naming convention, and it
sets and then resets the application_name GUC attached to every
existing role.  While we've not had complaints, these surely are
not good things to do within a production installation, and
regress.sgml pretty clearly implies that we won't do them.

Rather than lose test coverage altogether, let's just move this
script somewhere where it will get run by "make check" but not
"make installcheck".  src/test/modules/ already has that property.

Since it seems likely that we'll want other regression tests in
future that also exceed the constraints of "make installcheck",
create a generically-named src/test/modules/unsafe_tests/
directory to hold them.

Discussion: https://postgr.es/m/16638.1468620817@sss.pgh.pa.us

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/c91504b958e1647f87ffb7168e61db75ca09b928

Modified Files
--------------
src/test/modules/Makefile                                  |  1 +
src/test/modules/unsafe_tests/.gitignore                   |  4 ++++
src/test/modules/unsafe_tests/Makefile                     | 14 ++++++++++++++
src/test/modules/unsafe_tests/README                       |  8 ++++++++
.../unsafe_tests}/expected/rolenames.out                   |  0
.../{regress => modules/unsafe_tests}/sql/rolenames.sql    |  0
src/test/regress/parallel_schedule                         |  2 +-
src/test/regress/serial_schedule                           |  1 -
8 files changed, 28 insertions(+), 2 deletions(-)


pgsql-committers by date:

Previous
From: Peter Eisentraut
Date:
Subject: pgsql: Fix breakage introduced in pg_lsn_in()
Next
From: Tom Lane
Date:
Subject: pgsql: Blind attempt to fix SSPI-auth case in 010_dump_connstr.pl.