On 11/30/18 12:35 AM, Michael Paquier wrote:
On Fri, Nov 30, 2018 at 01:36:43AM +0000, Michael Paquier wrote:
Switch pg_verify_checksums back to a blacklist
This basically reverts commit d55241af705667d4503638e3f77d3689fd6be31,
leaving around a portion of the regression tests still adapted with
empty relation files, and corrupted cases. This is also proving to be
failing to check properly relation files located in a non-default
tablespace path.
So, jacana is not in love with this commit for the tests:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=jacana&dt=2018-11-30%2003%3A34%3A23
The error is here:
server started
# Postmaster PID for node "node_checksum" is 14820
error running SQL: 'psql:<stdin>:1: ERROR: directory
"/home/pgrunner/bf/root/HEAD/pgsql.build/src/bin/pg_verify_checksums/tmp_check/t_002_actions_node_checksum_data/ts_corrupt_dir"
does not exist'
What visibly happens is that we create a path for a tablespace which
will include a corrupted table. However psql tries to run CREATE
TABLESPACE but gets back a complain about the path not existing. The
thing is that mkdir has been used beforehand so the path should be
here.
I have tested this set of TAP tests on my Windows 10 VM with MSVC 2015
before commit and those tests passed. Could it be that jacana is using
msys, so a trick similar to what is used in 014_unlogged_reinit.pl is
necessary?
Andrew, could you confirm if the patch attached works? I would prefer
not doing a blind fix if possible..
All it actually needs is this additional line after the mkdir:
$tablespace_dir = TestLib::real_dir($tablespace_dir);
Explanation: TAP tests on msys need to run with the DTK perl, which
understands msys virtualized paths. Postgres, however, does not
understand such paths, so before you use such a value in, say, CREATE
TABLESPACE, you need to translate it into a path on the underlying file
system.
cheers
andrew
--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
