Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> Oh, what I normally do is to look at regression.diff, and if that looks
> OK, I just apply it to the expected file like this:
> cd expected
> patch < ../regression.diff
Oh, that explains a few things ...
It's much better to just copy the result file over the expected file
once you've decided it's OK. The regression.diff file is inexact
because of the diff switches that are used.
If you need to update expected variants your machine doesn't generate,
make a fresh regular diff off the expected and actual, and apply that
to the other variants.
regards, tom lane