Re: Can we clarify the purpose of `git diff -s`?

> Junio C Hamano <gitster@pobox.com> writes:
> 
> > Sergey Organov <sorganov@gmail.com> writes:
> >
> >> I entirely agree with your conclusion: obviously, -s (--silent) and
> >> --no-patch are to be different for UI to be even remotely intuitive, and
> >> I'd vote for immediate fix of --no-patch semantics even though it's a
> >> backward-incompatible change.
> >
> > I forgot to write about this part.
> >
> > tl;dr.  While I do not think the current "--no-patch" that turns off
> > things other than "--patch" is intuitive, an "immediate" change is
> > not possible.  Let's do one fix at a time.
> >
> > The behaviour came in the v1.8.4 days with a series that was merged
> > by e2ecd252 (Merge branch 'mm/diff-no-patch-synonym-to-s',
> > 2013-07-22), which
> >
> >  * made "--no-patch" a synonym for "-s";
> >
> >  * fixed "-s --patch", in which the effect of "-s" got stuck and did
> >    not allow the patch output to be re-enabled again with "--patch";
> >
> >  * updated documentation to explain "--no-patch" as a synonym for
> >    "-s".
> >
> > While it is very clear that the intent of the author was to make it
> > a synonym for "-s" and not a "feature-wise enable/disable" option,
> > that is what we've run with for the past 10 years.  You identified
> > bugs related the "-s got stuck" problem and we recently fixed that.
> 
> I wonder, why this intention of the author has not been opposed at that
> time is beyond my understanding, sorry! What exactly did it bring to
> make --no-patch a synonym for -s? Not only it's illogical, it's even not
> useful as being more lengthy.

> > But "git diff --stat --patch --no-patch", which suddenly starts
> > showing diffstat after you make "--no-patch" no longer a synonym for
> > "-s", has a much larger potential to break the existing workflows.
> > And I do not think asking the users who followed the documented
> > "--no-patch is a synonym for -s" to change their script because we
> > decided to make "--no-patch" no longer a synonym is much harder.
> 
> Why somebody would use --no-patch instead of -s when she means -s? Is't
> it obvious that
> 
>    git diff --stat --patch -s
> 
> is not only shorter but dramatically more clear than
> 
>    git diff --stat --patch --no-patch
> 
> ???
> 
> Taking this into account, I'd predict no breakage at all.

> > So, no, I do not think we can immediately "fix".  I do not think
> > anybody knows if it can be done "immediately" or needs a careful
> > planning to transition, and I offhand do not know if it is even
> > possible to transition without fallout.
> 
> This has been expected, and this is one of the things that stops me from
> trying to "fix" anything in the Git UI recently. I think that perfectly
> legit carefulness from the maintainer to be conservative in accepting of
> such changes goes a bit too far, sorry!