bug#35624: log-view-diff regression

[Dmitry Gutov]

On 16.05.2019 23:04, Juri Linkov wrote:

> > Was it back in the days of CVS?
>
> Nowadays I'm using it every day on Git: first I make the log buffer
> writable by typing `C-x C-q', then type `C-o' at the top of the log
> buffer to insert an empty line to make this feature available.

Interesting. I could probably just use Git from console for this, though.

> > Could you explain what are you using it for? I understand a diff between
> > revisions, and I understand a diff of the working tree against the index,
> > but this kind of a diff seems pretty unusual.
>
> The last time I used this feature today: I needed to revert some
> previous commits partially, and before committing the reverting changes
> I needed to check if necessary changes were reverted.

OK. But just for the sake of the argument, here's how you could do the 
same thing:

1. You do the "partially revert" commit, then check whether it does what 
you want it to do without having to insert the extra line.
2. If it doesn't, you make the necessary extra change and do an amend 
commit (C-x v v; C-x C-e; C-c C-c), then goto 1. Otherwise end scenario.

> > I have tried to imagine using it, but it's still hard. Maybe doing it once
> > a few days to compare the current progress against the master
> > branch.
>
> Yes, I believe a try is worth a thousand words.

I actually meant that it's hard for me to imagine really needing this. 
Or doing it more often than once in a several days.

And I do sometimes revert parts of commits, usually doing something 
other than what you explained.

> > Though C-u C-x v d master RET C-j might be faster that looking for
> > the revision that corresponds to the master branch in the log.
>
> The log buffer helps to select the revisions to compare
> because it displays dates and summaries of commits.

How about 'C-u d', then? Like I suggested in the other email.