Re: [HACKERS] A note about debugging TAP failures

On 4/25/17 11:00, Daniel Gustafsson wrote:
> Makes sense, assuming a “clean slate” to run on seems a reasonable assumption
> for the test and it makes for simpler code in PostgresNode.  Updated the patch
> which now have static datadir names; retains on PG_TESTS_NOCLEAN or any kind of
> test failure; and allows for cleaning datadirs without having to use make clean
> (seems the wrong size hammer for that nail).

Committed.  I had to make some changes in the pg_rewind tests.  Those
make data directories with conflicting names, which did not work anymore
because the names are no longer random.  Other than that this is pretty
nice.  Thanks!

-- 
Peter Eisentraut              http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services