Re: js/difftool-no-index, was Re: What's cooking in git.git (May 2019, #02; Tue, 14)

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: May 15, 2019, 08:24 UTC
Message-ID: <nycvar.QRO.7.76.6.1905151015540.44@tvgsbejvaqbjf.bet>
In-Reply-To: <xmqqef508q0y.fsf@gitster-ct.c.googlers.com>

Hi Junio,

On Wed, 15 May 2019, Junio C Hamano wrote:

Show 19 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
>
> > On Tue, 14 May 2019, Junio C Hamano wrote:
> >
> >> * js/difftool-no-index (2019-05-09) 1 commit
> >>  - difftool --no-index: error out on --dir-diff (and don't crash)
> >>
> >>  The "--dir-diff" mode of "git difftool" is not useful in "--no-index"
> >>  mode; they are now explicitly marked as mutually incompatible.
> >>
> >>  Will merge to 'next'.
> >
> > Seeing as the following is a part of v2.22.0-rc0's release notes:
> >
> > 	 * "git difftool" can now run outside a repository.
> >
> > maybe we could get this to `master` instead of `next`?
>
> There is no 'instead of'.  Topics advance one step at a time.

Well, some topics seem to enter `next` via `master`...

Show 6 quoted lines

> I do not recall seeing you respond to my review comments on the
> "error out" patch <xmqqv9ykdr1w.fsf@gitster-ct.c.googlers.com>, but
> thinking about it further, assuming that my rephrasing of your
> justification in my review was correct, I wonder if it is even more
> correct if we do not error out but simply behave as if "--dir-diff"
> option were not given.

I misunderstood that to be a rhetoric question, like in the "thinking out loud" you favor. Sorry.

Show 6 quoted lines

> That is, if there were some way to distinguish tracked and untracked
> paths in --no-index mode, --dir-diff would copy only the tracked parts
> to temporary directories and run tree comparison.  Since there is no
> "untracked paths" by definition in --no-index mode, we do not even
> have to make copies but just compare the two directories given using
> the same tree comparison tool.

As far as `--no-index` is concerned, all files are untracked anyway. There is no index, so there are no staged/committed/tracked files.

> Wouldn't that be more useful behaviour than telling the end users to
> run the backend tree comparison tool themselves?

I am lost. What behavior are you talking about? I *think* you mean to say that the `--dir-diff` mode under `--no-index` should essentially do what `--no-dir-diff` does? I think that's not the way I want to go. The `--dir-diff` mode only makes sense when there is a mix of tracked/untracked files, which simply is not the case for `--no-index`. And we should not make it "fall back" to its opposite. The user asked for `--dir-diff` and it just does not make sense.

In any case, the patch series I'd like to see advanced to `master` is not about supporting or not supporting `--dir-diff` with `--no-index`, but about preventing a segmentation fault when a user calls the (non-sense) combination of `--no-index` and `--dir-diff`.

Ciao, Dscho