From: Johannes Sixt <j6t@kdbg.org>
Date: Sat, 15 Jan 2022 15:38:08 GMT
Subject: Re: A new feature command request `cmd_ignore`
Message-ID: <9a7eac53-79fe-24f5-225a-d00d26e67259@kdbg.org>
In-Reply-To: <9a3837b4-8c67-e114-e977-f6c7b729ba45@gmail.com>

Am 15.01.22 um 11:45 schrieb Jaydeep Das:
> 
>> I think that is a paraphrase of the suggestion to
> 
>    git rm --cached settings
>    echo settings >> .gitignore
> 
>> when changes to a file 'settings' that was already tracked (and often
>> contains user-specific settings) should be ignored. This misguided and
>> short-sighted "solution" is repeated numerous times on Stackoverflow.
>> Not something that we should encourage.
> 
> So what should be done in this scenario? And why is it a discouraged and
> misguided thing?

It is misguided because if you commit the removal of 'settings', others
who pull/merge this change will either see their 'settings' be removed
(when they did not change the file) or their pull/merge will fail with
an error about a modified file being in the way (when they did modify
it). It causes a whole lot of grief for others.

There are different ways to address this problem.

- One is to commit a 'settings.template' file that must be copied to
'settings' and is adjusted as necessary, but not `git add`ed.

- Another is to have a mechanism that can include local files that carry
the user-specific settings.

- Yet another is to generate the final file from a tracked template and
a local, untracked, file during the build process.

-- Hannes