Hi all,
While looking at the buildfarm and reviewing the old snapshot patch, I
noticed that isolation tests cannot run with MSVC scripts all the
time. Take for example that:
http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=currawong&dt=2016-03-24%2004%3A55%3A21
Which leads to the following funny thing in contrib-install-check-C:
============================================================
Checking test_decoding
(using postmaster on localhost, default port)
============== dropping database "contrib_regression" ==============
DROP DATABASE
============== creating database "contrib_regression" ==============
CREATE DATABASE
ALTER DATABASE
============== running regression test queries ==============
=====================All 0 tests passed.
=====================
The origin of this problem is that those vcregress.pl is not able to
fetch any test names, and creates an temporary instance for nothing.
In order to get to a cleaner situation, I propose the following:
- Fix the Makefile of test_decoding to define ISOLATION_OPTS and
REGRESS_OPTS so as the msvc scripts can fetch the options to run with
isolation_tester correctly, and switch REGRESSCHECKS to REGRESS, then
ISOLATIONCHECKS to ISOLATION.
- Extend vcregress.pl so as it is able to detect ISOLATION[_OPTS] to list tests
- Add more cruft in subdircheck@vcregress.pl to guess if pg_regress,
isolation_tester, or both are needed
For the purpose of test_decoding, we need to be able to run both
pg_regress and isolation_tester.
Now, not testing those things has little impact perhaps... But I'd
rather fix those problems, and at least making the Makefile of
test_decoding more consistent does not sound bad to me. Thoughts?
Regards,
--
Michael
