Re: What's cooking in git.git (May 2023, #04; Thu, 11)

From: Felipe Contreras <felipe.contreras@gmail.com>
Date: May 12, 2023, 01:36 UTC
Message-ID: <CAMP44s0GAiFhoHFo4Ex464utf2Jm1F-LEdX2PZdUGEdp5x6Liw@mail.gmail.com>
In-Reply-To: <xmqqo7mqs7rp.fsf@gitster.g>

Junio C Hamano wrote:

Show 9 quoted lines

> * ds/merge-tree-use-config (2023-05-10) 1 commit
>   (merged to 'next' on 2023-05-11 at e0dab53028)
>  + merge-tree: load default git config
>
>  Allow git forges to disable replace-refs feature while running "git
>  merge-tree".
>
>  Will merge to 'master'.
>  source: <pull.1530.git.1683745654800.gitgitgadget@gmail.com>

Why was this series merged after only 11 minutes of review window? Are patches from GitHub favored over all others?

As I explained here [1], this is clearly the wrong approach because it ends up calling git_default_config *twice*.

The simplest fix is [2], which simply shuffles one line of code.

The proper fix is to reorganize the way merge_recursive_config() works, so git_xmerge_config is called only once at the right time (before any option parsing). That not only fixes the reported bug, but also makes merge.conflictStyle work for both real merges and trivial merges, along with fixing other issues.

[1] https://lore.kernel.org/git/645d672447ebb_13d3fe294f@chronos.notmuch/ [2] https://lore.kernel.org/git/20230511215608.1297686-1-felipe.contreras@gmail.com/ [3] https://lore.kernel.org/git/20230511235951.2309462-1-felipe.contreras@gmail.com/

-- 
Felipe Contreras