Re: [PATCH] push: add config option to --force-with-lease by default.

From: Francesco Mazzoli <f@mazzo.li>
Date: Jul 5, 2017, 16:43 UTC
Message-ID: <EA42AB04-D538-4B4A-BC19-CAE9A80EEA76@mazzo.li>
In-Reply-To: <xmqqa84ikj61.fsf@gitster.mtv.corp.google.com>

Show 16 quoted lines

> On 5 Jul 2017, at 17:17, Junio C Hamano <gitster@pobox.com> wrote:
> 
> The take-away lesson that the earlier thread gave me was that the
> order in which the three options are ranked by their desirebility
> in the UI (and the order we would like to encourage users to use)
> is, from the most to the least preferrable:
> 
> - "--force-with-lease=<ref>:<expect>" that is safer than "--force";
> 
> - "--force" that is known to be dangerous, and does not pretend to
>   be anything but;
> 
> - "--force-with-lease" that pretends to be safer but is not.
> 
> The last form should eventually be eliminated, as there is no way to
> correctly intuit what the expected object should be.

What's not clear to me is what the intended workflow using `--force-with-lease=<ref>:<expect>` is. Intuitively it seems extremely cumbersome to manually pluck a revision each time, especially when dealing with commits that all have the same description.

On the other hand for my workflow `--force-with-lease` works quite well because I tend to use it in cases where me and a colleague are working on the same PR, and thus I'm not doing anything else (including fetching).

Moreover, it seems to me that the problem `--force-with-lease` is just one of marketing. `--force-with-lease` is strictly more "safe" than `--force` in the sense that it'll reject some pushes that `--force` will let through. I think that if we advertise it better including its drawbacks it can still be better than no checks at all.

Francesco