Documentation bug (?) when describing `zdiff3` merge format

The docs at https://git-scm.com/docs/git-merge#_how_conflicts_are_presented describe the following snippets in `diff3` and `zdiff3` style as equivalent. They do not seem equivalent to me, so either this is a mistake or the `zdiff3` style is counterintuitive needs a better explanation.

diff3 style:

```
Here are lines that are either unchanged from the common
ancestor, or cleanly resolved because only one side changed,
<<<<<<< yours:sample.txt
or cleanly resolved because both sides changed the same way.
Conflict resolution is hard;
let's go shopping.
||||||| base:sample.txt
or cleanly resolved because both sides changed identically.
Conflict resolution is hard.
=======
or cleanly resolved because both sides changed the same way.
Git makes conflict resolution easy.

And here is another line that is cleanly resolved or unmodified. ```

zdiff3 style:

```
Here are lines that are either unchanged from the common
ancestor, or cleanly resolved because only one side changed,
or cleanly resolved because both sides changed the same way.
<<<<<<< yours:sample.txt
Conflict resolution is hard;
let's go shopping.
||||||| base:sample.txt
or cleanly resolved because both sides changed identically.
Conflict resolution is hard.
=======
Git makes conflict resolution easy.

And here is another line that is cleanly resolved or unmodified. ```

The problem is that, I believe, the "or cleanly resolved because both sides changed identically." sentence should not be part of the **base** in the latter example, since that whole line was moved outside the conflict.

I'd appreciate knowing which it is.

        Thanks,
             Ilya

Am 08.08.24 um 03:22 schrieb punk.lion0906@fastmail.com:

I don't think that the documentation wants to claim any equivalence. I can only see it says that "diff3 shows the conflict like this", and "zdiff3 shows the conflict like that".

This is exactly the problem that zdiff3 has: There is no way to indicate that some line in the common ancester was resolved identically and is now outside the conflict markers. Do not use zdiff3 if you cannot live with this deficiency. This is in fact not made explicit in the documentation.

-- Hannes

Johannes Sixt <j6t@kdbg.org> writes:

Yes, indeed that is a problem and that is not I personally do not touch it even with a 10ft pole. We might want to stress it a bit more to warn potential users in our documentation.

#leftoverbits then?

Thanks.

On Wed, Aug 7, 2024 at 6:22 PM <punk.lion0906@fastmail.com> wrote:

This line _still changed_, even though there were conflicts around it, meaning that if we didn't have the "... identically" line in base, we wouldn't see that this line was removed anywhere in the diff. I agree with the other responders that this is awkward, as it means you can't reconstruct the original 'base' version from these diff markers; if you tried, you'd get a base version that has "or cleanly ... the same way" followed by "or cleanly ... identically". For manual conflict resolution where you generally won't ever keep 'base' or attempt to reconstruct it, however, it seems fine to me if you're aware of it happening.

Ah, thanks for pointing this out, I absolutely missed that. My mind parsed "... identically" and "... the same way" in the same way.

Now the situation makes a bit more sense, I'm getting a sense of *what* `zdiff3` is doing, even though I still don't quite follow the why.

In any case, in my mind this is now a documentation bug where `zdiff3` is not explained as well as it could be. I don't understand the situation enough to provide the fix, though.

Thank you, everyone, for the help!
               Ilya.