On 2014-10-19 12:38:52 -0400, Tom Lane wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
> > I still think that reducing the need for having to do this is a good
> > idea, having to manually edit regression output to add regexes will be a
> > PITA.
>
> Yeah :-(.
With regard to what triggered this subthread, I still think we should
remove more irreproducible output to avoid having to do that more
often...But since I also think regexes would really be useful for other
cases, that doesn't preclude going forward with regex diffs either.
> We already have two features somewhat related to this, viz
> prototype expected files in output/ and multiple expected files. And
> each of them is a pain when it comes time to update the expected output.
> So that analogy is damping my enthusiasm a bit. Still, it beats not
> being able to make a test at all. And to the extent that we could get
> rid of variant expected files, we could eliminate a class of committer
> mistakes that's bitten pretty much everyone at one time or another...
We probably can evn get rid of a fair bit of the output/ cases. The
input files still have to be generated, but it's still less work.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services
