Re: [PATCH 1/2] rebase: add a --rebase-merges=drop option

From: Philip Oakley <philipoakley@iee.email>
Date: Feb 21, 2023, 16:08 UTC
Message-ID: <a856dd16-9876-509b-6a99-11ea0020633c@iee.email>
In-Reply-To: <xmqqr0ukggk5.fsf@gitster.g>

Hi Junio,

On 20/02/2023 21:42, Junio C Hamano wrote:

Show 5 quoted lines

> Alex Henrie <alexhenrie24@gmail.com> writes:
>
>> Name the new option "drop" intead of "no" or "false" to avoid confusion
> This is traditionally called "flattening the history".  Don't we
> confuse uesrs by introducing a new phrase?

While "flatten.." is used on list, we rarely mention it in our man pages, and usually only in a cautionary manner via the rev-list-options.txt under "--show-linear-break".

It's not always clear what is meant by 'flattening' and which aspects are included/excluded from the flattened display. I suspect that a recent question on the git-users list [1] originates from the same confusions.

Maybe it's something that could be included in the Glossary to supplement the not well known how-to discussion in keep-canonical-history-correct.txt

Show 16 quoted lines

>
> rebase-merges is about transplanting the history without flattening,
> i.e. keeping the mergy commit graph topology.  If there are only two
> kinds of rebase (i.e. keeping the topology which is rebase-merges
> and the other "flattening" kind) operation, shouldn't the option be
> called "--no-rebase-merges" instead?  --rebase-merges=no is also
> understandable.
>
>> in the future if --rebase-merges grows the ability to truly "rebase"
>> merge commits by reusing the conflict resolution information from the
>> original merge commit, and we want to add an option to ignore the
>> conflict resolution information.
> I am not sure why such a change "in the future" is not merely a
> bugfix of the current "--rebase-merges", though.  Once it is fixed,
> is there a reason to make the fixed behaviour only available behind
> an option?

[1] https://groups.google.com/d/msgid/git-users/057bd9e2-b20b-4794-b8a0-bc16ede374c1n%40googlegroups.com

--

Philip