Re: [PATCH v3] config.c: split some variables to $GIT_DIR/config.worktree

[Duy Nguyen <pclouds@gmail.com>]

On Thu, Apr 2, 2015 at 3:56 AM, Max Kirillov <max@max630.net> wrote:
> On Tue, Mar 31, 2015 at 07:14:39PM +0700, Nguyễn Thái Ngọc Duy wrote:
>>  The general principle is like in the last mail: .git/config is for
>>  both shared and private keys of main worktree (i.e. nothing is
>>  changed from today).  .git/worktrees/xx/config.worktree is for
>>  private keys only (and private keys in .git/config are ignored)
>>
>>  With this we don't have to bump core.repository_format_version for
>>  main worktree because nothing is changed. There will be problems
>>  with info/config.worktree:
>>
>>   - it's customizable, so expect the user to break it (*)
>
> I would rather say it's manual tuning rather than a break.

But if the user accidentally deletes core.worktree then something will break.

>>   - if we add new stuff to the template, we'll need to help migrate
>>     current info/core.worktree (which does not have new stuff).
>>     Auto updating this file could be risky. I'm tend to just
>>     warn the user that this and that keys should be included and let
>>     them modify the file.
>
> I don't think there even should be warning. Just don't
> change the info/config.worktree in the existing repositories
> and let users extend it as they feel a need for it.
>
> Later there could be a tool (an option to git config for
> example) which adds a variable or pattern to the
> info/core.worktree and copied existing variable(s) from
> commong config to worktree-specific ones.

I was thinking "git checkout --to" would do this. It checks if the
main worktree has info/core.worktree, if not, re-init the repo to add
this file from default template.

>> @@ -1932,6 +2002,23 @@ int git_config_set_multivar_in_file(const char *config_filename,
>>
>>       store.multi_replace = multi_replace;
>>
>> +     if (git_common_dir_env && is_config_local(key)) {
>> +             if (!config_filename)
>> +                     config_filename = filename_buf = git_pathdup("config.worktree");
>> +             /* cheap trick, but should work 90% of time */
>> +             else if (!ends_with(config_filename, ".worktree"))
>> +                     die("%s can only be stored in %s",
>> +                         key, git_path("config.worktree"));
>
> Is `config_filename` set only for cases when config file is
> explicitly set for "git config" command with "--file"
> option? Then probably here should not be any intelligence at
> all; if user resort to manual picking the file he must be
> allowed to do stupid things.

Yes. It may make sense in the previous versions where this feature
could work in a normal worktree, but when it becomes
multiworktree-specific, I guess we can drop this.
-- 
Duy