Bruce Momjian writes:
> Amazing you find 688 bytes worth discussing. I know you said "what
> happens if everyone adds their scripts", but something that would be a
> mess if everyone did it isn't always a proper way to judge if something
> is appropriate.
I said, if everyone adds their test methodologies. That leads to
discrepancies, more of them down the road if one method changes and the
other doesn't catch up. For instance, your method just calls pg_ctl,
createdb, etc. from the path. If people already have a stable
installation of PostgreSQL on their machine, then this will test the wrong
installation. So, from now on, if someone submits a test result I have to
ask, "which method did you use" -- "don't use that method, because it's
wrong". That is one instance, and I'm sure you'll fix it, but there might
be more. What I'm saying is, we were in a discussion about improving the
testing of PostgreSQL, and this is not a step forward. If we need to
improve the testing mechanisms for various purposes -- patch application,
automated testing, etc. -- let's look at it and see how we can improve the
current infrastructure without inventing a parallel one. At this point,
I'm not sure why "make check" doesn't serve you. Perhaps you are not
fully aware of what it does (I guess so, from looking at your script).
--
Peter Eisentraut peter_e@gmx.net