Thread: Regression tests for preload extension

Regression tests for preload extension

From
"Kevin Grittner"
Date:
I wrote a little extension to store a few small strings in shared
memory.  It seems to be working fine, and now I would like to write
some regression tests; but it's not immediately obvious to me how I
can do that.  The approach used by, for example, citext doesn't
work, because I don't see how to set shared_preload_libraries for
the server startup.  The existing contrib extensions which preload
either seem to do something ad hoc or skip regression tests
entirely, so I suspect that is my choice; but I figured I should
ask.
-Kevin


Re: Regression tests for preload extension

From
Tom Lane
Date:
"Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
> I wrote a little extension to store a few small strings in shared
> memory.  It seems to be working fine, and now I would like to write
> some regression tests; but it's not immediately obvious to me how I
> can do that.  The approach used by, for example, citext doesn't
> work, because I don't see how to set shared_preload_libraries for
> the server startup.  The existing contrib extensions which preload
> either seem to do something ad hoc or skip regression tests
> entirely, so I suspect that is my choice; but I figured I should
> ask.

Hm.  pg_regress.c goes to some trouble to allow you to set session-level
options by setting PGOPTIONS in its environment, but that won't work
for options that have to be given to the postmaster.  Maybe we should
invent a pg_regress switch that allows additional switches to be given
to the temp postmaster.  Of course, this is never gonna work for
"make installcheck".
        regards, tom lane