Re: [PATCH 1/4] git-gc --auto: add pre-auto-gc hook

From: Junio C Hamano <gitster@pobox.com>
Date: Apr 1, 2008, 04:51 UTC
Message-ID: <7vsky6jhz3.fsf@gitster.siamese.dyndns.org>

Miklos Vajna <vmiklos@frugalware.org> writes:

Show 23 quoted lines

> If such a hook is available and exits with a non-zero status, then
> git-gc --auto won't run.
>
> Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
> ---
>  builtin-gc.c |   22 ++++++++++++++++++++++
>  1 files changed, 22 insertions(+), 0 deletions(-)
>
> diff --git a/builtin-gc.c b/builtin-gc.c
> index 8cef36f..acd63be 100644
> --- a/builtin-gc.c
> +++ b/builtin-gc.c
> @@ -157,6 +157,25 @@ static int too_many_packs(void)
>  	return gc_auto_pack_limit <= cnt;
>  }
>  
> +static int run_hook()
> +{
> +	const char *argv[2];
> +	struct child_process hook;
> +
> +	argv[0] = git_path("hooks/pre-auto-gc");
> +	argv[1] = NULL;

Hmm. I wonder if the hook wants any parameters, even though I do not think of any offhand.

Show 7 quoted lines

> +	if (access(argv[0], X_OK) < 0)
> +		return 0;
> +
> +	memset(&hook, 0, sizeof(hook));
> +	hook.argv = argv;
> +	hook.no_stdin = 1;
> +	hook.stdout_to_stderr = 1;

I understand no_stdin, but is there a reason to do stdout_to_stderr?

> +	return run_command(&hook);
> +}

Don't we want to distinguish between the case where start_command() failed, wait_or_whine() failed on waitpid(), the command was killed with signal, or the command actually ran correctly and decided that you should not run "git gc --auto" by exiting non-zero?

I think it is prudent to refrain from running "git gc --auto" in any of the failure cases I listed above, but shouldn't the cases other than the last one at least issue a warning?