Re: [PATCH 1/2] builtin/rebase.c: Emit warning when rebasing without a forkpoint

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 31, 2023, 21:52 UTC
Message-ID: <xmqq1qfiubg5.fsf@gitster.g>
In-Reply-To: <xmqqbkenszfa.fsf@gitster.g>

Junio C Hamano <gitster@pobox.com> writes:

Show 6 quoted lines

> I am not commenting on the tests, as the above code probably needs
> to be corrected first so that folks who want to squelch the message
> and want the "forkpoint behaviour by default when rebuilding on the
> usual upstream" behaviour can do so by setting the variable to true.
>
> And that obviously need to be tested, too.

Another worrysome thing about rebase.forkpoint is that it will be inevitable for folks to start complaining that it does not work the way other configuration variables do. Setting the variable to 'true' is not the same as passing '--fork-point=true' from the command line.

I actually think it would be a lot larger behaviour change with a huge potential to be received as a regression if we start making the variable to mean the same thing as passing '--fork-point=true'. People may like the current "if you are rebuilding your branch on its usual upstream, pay attention to the rebase and rewind of the upstream itself, but if you are giving an explicit upstream from the command line, the tool does not second guess you with the fork-point heuristics" behaviour and prefer to set it to true. We would be breaking them big time if suddenly the rebase.forkpoint=true they set previously starts triggering the fork-point heuristics when they run "git rebase upstream". So that needs to be kept in mind when/if we fix the "setting the variable, even to 'true', will squelch the warning".