David Rowley <dgrowleyml@gmail.com> writes:
> On Sat, Oct 18, 2014 at 1:39 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Not sure how we get there without writing our own diff engine though :-(.
(Note that after a bit of looking around, it seems like it might not be
that hard to do something like this in Perl. Perl is already
nearly-required for building from source, could we require it for running
the regression tests? Though we'd also need you to install
Algorithm::Diff or suchlike, which I think isn't in a basic Perl install.)
> I had imagined that we wouldn't need this, but perhaps my workflow is just
> different from yours. When I make changes which make tests fail for a valid
> reason I'd use beyondcompare to cherrypick the actual back into the
> expected, but I suppose others might just apply the diff into the
> expected.... Umm, but then wouldn't you just copy the whole actual file
> over to expected?
That's what I usually do, except when dealing with the ones that are
generated from output/ files. (Which are a PITA to update, and I guess
files containing wildcard matches would be too.)
> So why do we need diffs? Couldn't this matching tool
> just report where the first non-matching line appeared in the file?
Not too helpful for buildfarm reports. Nor for anyone else, really;
you're just pushing the problem of identifying the important difference
back onto the user.
regards, tom lane