Re: [PATCH 1/7] "git" calls help_unknown_cmd(""); "git help" and "git help -a" return 0

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

Scott R Parish <srp@srparish.net> writes:

> Signed-off-by: Scott R Parish <srp@srparish.net>
> ---
>  git.c  |    5 ++---
>  help.c |    4 ++--
>  2 files changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/git.c b/git.c
> index 853e66c..e1c99e3 100644
> --- a/git.c
> +++ b/git.c
> @@ -445,9 +445,8 @@ int main(int argc, const char **argv)
>  		if (!prefixcmp(argv[0], "--"))
>  			argv[0] += 2;
>  	} else {
> -		/* Default command: "help" */
> -		argv[0] = "help";
> -		argc = 1;
> +		/* The user didn't specify a command; give them help */
> +		help_unknown_cmd("");

Sorry, but I fail to see why this is an improvement.

If you type "git<Enter>", before this patch we call cmd_help()
without help_cmd, which gives the usage string and lists common
commands, and exits with 1.

With this patch, you get "git: '' is not a git-command",
followed by list of common commands, and then the invocation
exits with 1.

I think "git help" should exit with 0.  The user asked for help,
we give help, everything is going as expected and there is no
reason to exit with non-zero.

However, with the current implementation, these changes to
help.c also make "git<Enter>" to exit with 0 after it gives
help, which is not so nice (both "cvs" and "svn" without
parameter seem to exit with 1 and it sort of makes sense,
although I do not think it matters much).  As a few datapoints,
"cvs -H" exits with 1 and "svn help" exits with 0.

So in short,

 - "git" should retain the current behaviour (both output and
   exit code).

 - "git help" should retain the current output but probably
   should exit with 0.

 - Ditto for "git help -a".