Re: master 4803fba487 1/2: 'C-x v v' on a diff buffer commits it as a patch (bug#52349)

[Dmitry Gutov]

On 01.09.2022 14:48, Dmitry Gutov wrote:
> On 01.09.2022 14:45, Lars Ingebrigtsen wrote:
> > Dmitry Gutov<dgutov@yandex.ru>  writes:
> >
> > > The step of "applying the saved patch" relies on our Lisp code being
> > > correct and being able to handle any diff. It's not an outrageous
> > > assumption, but an assumption nevertheless.
> > Oh, I was assuming that we were assuming that we had "patch" installed
> > for this to work.  I don't think that's unreasonable for new
> > functionality like this.
>
> I suggested this upthread but didn't get any affirmative response.
>
> This would be easier, yes.

So... I ended up using 'patch', but not for backing up existing 
uncommitted changes.

It's the same reason as why 'hg shelve' and 'git stash' are a poor fit: 
when we try to apply the patch at the end, we can get a conflict (and 
not every such conflict will be easy enough to fix by hand).

The file-copying approach matches Git's implementation well, and 
hopefully will work with various backends, including CVS. I only tested 
with Hg for now.

Feedback welcome.