Re: unified vs context diffs (was Re: Strange Windows problem, lock_timeout test request)

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > * Robert Haas (robertmhaas(at)gmail(dot)com) wrote:
> >> True, but I'm with Heikki: it's a pedantic and unhelpful guideline.
>
> > Then let's change it, drop the preference, and update the documentation.
>
> I think we should drop the hard requirement for context-format, and
> instead say that it must not be plain (context-free) diff, since that
> clearly *is* a hard requirement.

Alright, I'll start making those changes.

> However, I liked the upthread suggestion (I think it was from Heikki)
> that we recommend that submitters actually take a moment to think about
> which format is more readable for their particular patch. Readability
> is important not only to help people who just give the patch a quick
> eyeball, but also to help the inevitable situations where hunks have
> to be applied by hand because the underlying code has changed. The
> less readable the patch, the more likely an error in doing that.
> (And I trust we've all learned by now that git isn't so good at merging
> that this isn't a problem.)

I'll include that point in my changes but I consider the chances of that
actually happening with any regularity to be essentially zero. Reducing
our requirement to a level where the default passes means that nearly
everyone, who hasn't already changed things to use a non-default
automatically, is going to just use the default.

Thanks,

Stephen