Re: [PATCH v2 1/2] branch: refactor "edit_description" code path

[Junio C Hamano <gitster@pobox.com>]

Rubén Justo <rjusto@gmail.com> writes:

> Minor refactoring to reduce the number of returns in the switch case
> handling the "edit_description" option, so the calls to strbuf_release
> can also be reduced.  New resources to be added also do not need to be
> released in multiple places.
>
> Signed-off-by: Rubén Justo <rjusto@gmail.com>
> ---
>  builtin/branch.c | 17 ++++++++---------
>  1 file changed, 8 insertions(+), 9 deletions(-)
>
> diff --git a/builtin/branch.c b/builtin/branch.c
> index 55cd9a6e99..5229cb796f 100644
> --- a/builtin/branch.c
> +++ b/builtin/branch.c
> @@ -614,7 +614,7 @@ static int edit_branch_description(const char *branch_name)
>  	strbuf_reset(&buf);
>  	if (launch_editor(edit_description(), &buf, NULL)) {
>  		strbuf_release(&buf);
> -		return -1;
> +		return 1;
>  	}
>  	strbuf_stripspace(&buf, 1);


Our API convention is to signal a failure with negative return
value.  Granted that this is not a general API but is merely a
helper function in the implementation of a single command, it would
be less confusing if you sticked to the convention.

Unless there is a compelling reason not to, that is.

> @@ -791,6 +791,7 @@ int cmd_branch(int argc, const char **argv, const char *prefix)
>  	} else if (edit_description) {
>  		const char *branch_name;
>  		struct strbuf branch_ref = STRBUF_INIT;
> +		int ret = 0;
>  
>  		if (!argc) {
>  			if (filter.detached)
> @@ -803,19 +804,17 @@ int cmd_branch(int argc, const char **argv, const char *prefix)
>  
>  		strbuf_addf(&branch_ref, "refs/heads/%s", branch_name);
>  		if (!ref_exists(branch_ref.buf)) {
> -			strbuf_release(&branch_ref);
> -
>  			if (!argc)
> -				return error(_("No commit on branch '%s' yet."),
> +				ret = error(_("No commit on branch '%s' yet."),
>  					     branch_name);
>  			else
> -				return error(_("No branch named '%s'."),
> +				ret = error(_("No branch named '%s'."),
>  					     branch_name);

OK.  These are good uses of a new variable 'ret'.  Note that error()
returns negative one.

> -		}
> -		strbuf_release(&branch_ref);
> +		} else
> +			ret = edit_branch_description(branch_name);
>  
> -		if (edit_branch_description(branch_name))
> -			return 1;
> +		strbuf_release(&branch_ref);
> +		return ret;

When editor failed, we leaked branch_ref strbuf, but we no longer
do.

Which is good.

This makes cmd_branch() return -1 (when we see error() call) or 1
(when edit_branch_description() fails and returns 1).  I would
suggest to

 * Fix the return value of edit_branch_description() so that it
   signals a failure by returning -1

 * cmd_branch() to return (or call exit() with) -ret, as ret has 0
   when everything is peachy, and negative in any error code paths.