Re: support for git commit --amend/--signoff

[Stefan Monnier]

>>> Can you please show how this is better than adding a single argument
>>> to the vc-checkin method?  (For which the code already exists).
>> It makes more of the state plainly visible to the user, editable with
>> Emacs's usual commands, rather than hidden in variables that are much
>> more difficult for the user to control.
> How is it more difficult to control something that has an explicit
> command attached to it?  (that you've stated that we need to have
> anyway)

It's more difficult because you're limited to what the command lets
you do, whereas with plain text, you can do anything you like
(including running the provided command).

>> I.e. it's more Emacsy by keeping things shallow.
> That's not quite true for a lot of things we do in Emacs.

And in most cases, that is a problem.  Sometimes it's justified, and
sometimes it's just because noone has gone through the trouble to make
it better.

>> It means the commit doesn't amend.  Same thing if you set your vars
>> accidentally or if you call the toggle command accidentally, ...
>> 
>> I really don't see it as a problem.  Even if it's done accidentally,
>> it's obvious for the user what the behavior will be, since it's written
>> in plain text.

> Why deal with any potential complications that are not needed at all?

What complications are you referring to?

>>> --amend and --signoff simply do not fit the header paradigm.  
>>> Can we please admit that and move on?
>> I really don't see it.
> Are willing to implement your version and ask users if they prefer it?

Don't know.  But I don't want the implementation you propose.  I want
the information to be shallow and readily available for the user to
manipulate as she sees fit.


        Stefan