Re: [RFC] How to pass Git config command line instructions to Submodule commands?

[Lars Schneider <larsxschneider@gmail.com>]

> On 28 Apr 2016, at 13:25, Jeff King <peff@peff.net> wrote:
> 
> On Thu, Apr 28, 2016 at 01:06:45PM +0200, Lars Schneider wrote:
> 
>> I got my Git-LFS use case working with the patch below. 
>> For me it was necessary to export GIT_CONFIG_PARAMETERS
>> to make it available to the Git process if the process is 
>> invoked as follows [2]: 
>> 
>> (sanitize_submodule_env; cd "$sm_path" && git <something>")
> 
> Hrm. I'm not sure why you need to export. Or perhaps, I am not sure why
> it ever works in the first place in git-submodule.sh. In this code:
> 
>> diff --git a/git-submodule.sh b/git-submodule.sh
>> index 2a84d7e..b02f5b9 100755
>> --- a/git-submodule.sh
>> +++ b/git-submodule.sh
>> @@ -199,7 +199,7 @@ sanitize_submodule_env()
>> {
>> 	sanitized_config=$(git submodule--helper sanitize-config)
>> 	clear_local_git_env
>> -	GIT_CONFIG_PARAMETERS=$sanitized_config
>> +	export GIT_CONFIG_PARAMETERS=$sanitized_config
>> }
> 
> If you already have $GIT_CONFIG_PARAMETERS exported when we enter the
> function, then we should not need to re-export it when changing the
> value in the final line (the export bit is retained by the shell). But
> if you don't have it set already, then $sanitized_config must by
> definition be empty.
> 
> So it should do the right thing without the export.
> 
> At the same time, clear_local_git_env() will call "unset" on
> GIT_CONFIG_PARAMETERS. Which would clear the export bit, meaning the
> final line doesn't ever have any impact on sub-programs, and the whole
> thing is totally broken. But then, why does the test in t5550 pass?
> 
> Confused...

I am no expert in the Submodule code but I think the cloning of
the submodules is not yet guarded with sanitize_submodule_env [3].
That means the submodule is cloned with the GIT_CONFIG_PARAMETERS
of the super project. That might explain why t5550 passes as the 
credential config is only used in that area.

The submodule checkout is guarded with sanitize_submodule_env
and therefore my Git-LFS filter use case is affect.

Does this sound reasonable?

Thanks,
Lars

[3] https://github.com/git/git/blob/3ad15fd5e17bbb73fb1161ff4e9c3ed254d5b243/git-submodule.sh#L704-L711
[4] https://github.com/git/git/blob/3ad15fd5e17bbb73fb1161ff4e9c3ed254d5b243/git-submodule.sh#L811