David Fetter <david@fetter.org> writes:
> On Mon, Sep 27, 2010 at 03:11:07PM -0400, Robert Haas wrote:
>> Sure. But everyone using pg_regress will have to update their
>> regression test expected outputs.
> Again, I must be missing something super important. What is it that
> prevents people from doing
> find . -type f |xargs perl -pi.bak -e 's/\s+$//g'
I think the concern isn't about the change being hard to make, it's
about the fallout from having to have different expected-result files
for 9.1 versus previous releases. A lot of third-party modules try
to maintain source code compatibility across all PG releases they
support, and this would break that.
That's not necessarily a sufficient reason for us not to do it ---
but it's definitely not a negligible issue, either.
It should be noted that this won't be zero-cost for the core project
either. For example, any back-patched fix that adjusts regression test
outputs will have to deal with the incompatibility. And if we do put in
some git-level check on whitespace, it'll have to be made to only apply
to master not back branches.
regards, tom lane