Mark Bixby <mark@bixby.org> writes:
> Is it possible to re-run failing regression tests individually?
I believe so, but it's not very convenient in the "runcheck" mode, since
that normally wants to make a fresh install and start a temporary
postmaster. Instead, do a real install, start a real postmaster, and
do "make runtest" to create the regression DB in the real installation.
Then you can basically just do "psql regression <foo.sql" --- look at
the regression driver script to get the details of what switches to
pass and how to do the output comparison.
There are some order dependencies among the tests, but I think all the
ones you were having trouble with should be able to work this way in
an end-state regression DB.
Also, rerunning the whole suite is much quicker this way, since you
don't have to go through install/initdb/start postmaster each time.
BTW, the results you posted looked good --- with the exception of
plpgsql, the failing tests all seemed to be ones that are notorious
for platform-dependent output.
regards, tom lane