Re: [PATCH] builtin/maintenance: accept "none" as a maintenance strategy

Patrick Steinhardt <[email protected]> Tue, 4 Aug 2026 14:58:55 +0200
Newsgroups org.kernel.vger.git
Message-ID <[email protected]>
On Wed, Jul 29, 2026 at 03:40:06PM -0400, David Lin wrote:
> Commit d465be2327 (builtin/maintenance: don't silently ignore invalid
> strategy, 2025-10-24) changed scheduled maintenance to error on an
> unknown maintenance strategy instead of silently defaulting to the
> `none` strategy.
> 
> However, `parse_maintenance_strategy()` does not recognize `none`, so
> Git rejects a valid and documented strategy that can be used to override
> an existing strategy and disable maintenance tasks.

Oh, indeed.

> Accept `none` as a valid maintenance strategy and add tests to ensure
> it's accepted.

Makes sense. You can of course achieve the same thing by disabling
maintenance altogether, but it's a documented thing and users thus
rightfully expect the "none" strategy to exist.

> diff --git a/builtin/gc.c b/builtin/gc.c
> index 46999a99ab..3d1e39d46a 100644
> --- a/builtin/gc.c
> +++ b/builtin/gc.c
> @@ -1922,6 +1922,8 @@ static const struct maintenance_strategy geometric_strategy = {
>  
>  static struct maintenance_strategy parse_maintenance_strategy(const char *name)
>  {
> +	if (!strcasecmp(name, "none"))
> +		return none_strategy;
>  	if (!strcasecmp(name, "incremental"))
>  		return incremental_strategy;
>  	if (!strcasecmp(name, "gc"))

Yup, looks obviously correct.

> diff --git a/t/t7900-maintenance.sh b/t/t7900-maintenance.sh
> index a8d691719d..130c971b15 100755
> --- a/t/t7900-maintenance.sh
> +++ b/t/t7900-maintenance.sh
> @@ -1022,6 +1022,9 @@ test_expect_success 'maintenance.strategy is respected' '
>  		test_must_fail git -c maintenance.strategy=unknown maintenance run 2>err &&
>  		test_grep "unknown maintenance strategy: .unknown." err &&
>  
> +		test_strategy none </dev/null &&
> +		test_strategy none --schedule=weekly </dev/null &&
> +
>  		test_strategy incremental <<-\EOF &&
>  		git pack-refs --all --prune
>  		git reflog expire --all

And test looks obviously correct to me, too.

So other than Junio's remark about the SOB this patch looks good to me.
Thanks!

Patrick