Re: so/diff-merges-more (was Re: What's cooking in git.git (Feb 2023, #01; Thu, 2))

From: Glen Choo <chooglen@google.com>
Date: Mar 10, 2023, 21:19 UTC
Message-ID: <kl6la60kqp85.fsf@chooglen-macbookpro.roam.corp.google.com>
In-Reply-To: <878rg5vgvc.fsf@osv.gnss.ru>

Sergey Organov <sorganov@gmail.com> writes:

Show 19 quoted lines

> Glen Choo <chooglen@google.com> writes:
>
>> Sergey Organov <sorganov@gmail.com> writes:
>>
>>> I already agreed long ago that to be on the safe side we shouldn't
>>> simply change -m nowadays, and addressed that concern by putting -m
>>> behavior change under configuration option in the current series. So
>>> what's the reason of rejection?
>>
>> Before we start talking in circles again, I think the responses
>> elsewhere in the thread still accurately capture mine and Junio's
>> views, e.g.:
>>
>>   https://lore.kernel.org/git/xmqqedr28wwb.fsf@gitster.g
>> https://lore.kernel.org/git/kl6lr0v2i0gn.fsf@chooglen-macbookpro.roam.corp.google.com/
>
> There were no any response from you to my recent very direct question
> (that Junio did somewhat address), but what I figured from your and
> Junio explanation so far looks like:

Yes, thank you for summarizing the discussion. I think the following sums up my position quite accurately, and I found this very helpful.

> 1. The fact that -m does not imply -p is a mistake. There is no any
>    reasons this exact behavior could be useful. As such, it does not
>    make sense to support this exact behavior in --diff-merges. So the
>    reject of --diff-merges=[no-]hide.

Yes.

> 2. This mistake is too dangerous to fix due to subtle compatibility
>    problems, so we can't just fix -m behavior. Thus the reject of my
>    earlier patch "let -m imply -p".

Yes, unfortunately. If we didn't have compatibility problems, "-m imply -p" would be very good IMO. I am not that hopeful, but perhaps there are ways to get users to move off this compatibility problem. It would be something we would have to learn as a project though, it is not a problem unique to this series.

> 3. Moving behavior change under option is not worth it, as nobody
>    presumably needs this fixed -m behavior anyway (at least among 2
>    persons that are actually opposing the changes). So the reject
>    of "add diffMerges-m-imply-p configuration option" patch.

Yes, given how specific the option is (I think it would be the only config option that exists to change the behavior of a CLI option) I don't think the option is worth it in light of alternatives like a new "-d" flag or supporting "-mp".

> 4. Staring in the face inconsistency between -m and the rest
>    of short diff-merge options is not significant enough to reconsider
>    any of the above rejects.

Yes. Git carries many such inconsistencies.

> Did I outline your positions correctly and to the full extent? Anything
> else I forgot?

I cannot think of anything else. Thank you.

> I want to make sure I got it exactly right, as I'd still like to tweak
> my proposals, and I'd like to avoid the troubles I got into as much as
> possible.

Thanks. I hope that there is a good way forward for your proposals.

>
> Thanks,
> -- Sergey Organov