Re: [PATCH v2 2/3] config: unify code paths to get global config paths

From: Jeff King <peff@peff.net>
Date: Apr 9, 2021, 15:21 UTC
Message-ID: <YHBw7D/fEUdCckc8@coredump.intra.peff.net>
In-Reply-To: <dddc85bcf54e9b19f1612cf2a5be928dcb2bad7d.1617975637.git.ps@pks.im>

On Fri, Apr 09, 2021 at 03:43:25PM +0200, Patrick Steinhardt wrote:

Show 9 quoted lines

> There's two callsites which assemble global config paths, once in the
> config loading code and once in the git-config(1) builtin. We're about
> to implement a way to override global config paths via an environment
> variable which would require us to adjust both sites.
> 
> Unify both code paths into a single `git_global_config()` function which
> returns both paths for `~/.gitconfig` and the XDG config file. This will
> make the subsequent patch which introduces the new envvar easier to
> implement.

Seems like a good step forward. There is one minor issue with the implementation, though.

Show 13 quoted lines

> diff --git a/builtin/config.c b/builtin/config.c
> index 02ed0b3fe7..604a0973a5 100644
> --- a/builtin/config.c
> +++ b/builtin/config.c
> @@ -671,9 +671,9 @@ int cmd_config(int argc, const char **argv, const char *prefix)
>  	}
>  
>  	if (use_global_config) {
> -		char *user_config = expand_user_path("~/.gitconfig", 0);
> -		char *xdg_config = xdg_config_home("config");
> +		const char *user_config, *xdg_config;
>  
> +		git_global_config(&user_config, &xdg_config);

The pointer out-parameters make sense here, since we need to return two values. I notice they became const, so the function will hold on to ownership of the memory.

Show 9 quoted lines

> @@ -688,10 +688,8 @@ int cmd_config(int argc, const char **argv, const char *prefix)
>  		if (access_or_warn(user_config, R_OK, 0) &&
>  		    xdg_config && !access_or_warn(xdg_config, R_OK, 0)) {
>  			given_config_source.file = xdg_config;
> -			free(user_config);
>  		} else {
>  			given_config_source.file = user_config;
> -			free(xdg_config);
>  		}

...which is why we drop these free() calls. So far so good.

Show 12 quoted lines

> +void git_global_config(const char **user, const char **xdg)
> +{
> +	static const char *user_config, *xdg_config;
> +
> +	if (!user_config) {
> +		user_config = expand_user_path("~/.gitconfig", 0);
> +		xdg_config = xdg_config_home("config");
> +	}
> +
> +	*user = user_config;
> +	*xdg = xdg_config;
> +}

And here in the implementation we hold on to the static values forever.

I think your "did we initialize already" check isn't robust, though. expand_user_path() can return NULL, in which case every call would trigger a re-initialization (even leaking xdg_config if it was set in the last round).

So I think you'd need a separate "static int initialized" variable.

That said, I wonder if we should just pass ownership of the memory to the caller. It is a minor inconvenience that they will have to free() the result, but we're already doing that. And it removes any possibility of thread unsafety.

I guess it doesn't match git_system_config() as well, then. But arguably it should also just pass ownership (it also has only a handful of callers, and freeing the result would not be a big deal).

I'm OK with either solution, though.

-Peff