Re: [PATCH v3] gc: add `--expire-to` option

From: Junio C Hamano <gitster@pobox.com>
Date: Jan 16, 2025, 18:23 UTC
Message-ID: <xmqqzfjqr526.fsf@gitster.g>
In-Reply-To: <pull.1843.v3.git.1736994932003.gitgitgadget@gmail.com>

"ZheNing Hu via GitGitGadget" <gitgitgadget@gmail.com> writes:

Show 16 quoted lines

> From: ZheNing Hu <adlternative@gmail.com>
>
> This commit extends the functionality of `git gc`
> by adding a new option, `--expire-to=<dir>`. Previously,
> this feature was implemented in `git repack` (see 91badeb),
> allowing users to specify a directory where unreachable and
> expired cruft packs are stored during garbage collection.
> However, users had to run `git repack --cruft --expire-to=<dir>`
> followed by `git prune` to achieve similar results within `git gc`.
>
> By introducing `--expire-to=<dir>` directly into `git gc`,
> we simplify the process for users who wish to manage their
> repository's cleanup more efficiently. This change involves
> passing the `--expire-to=<dir>` parameter through to `git repack`,
> making it easier for users to set up a backup location for cruft
> packs that will be pruned.

Today I do not have enough time to do my usual commit log message critique. Please use "git show -s --format=reference" when referring to an earlier commit.

> Note: When git-gc is used with both `--cruft` and `--expire-to`,
> it does not pass `-a` to git-repack to delete all unreachable
> objects as `git gc --prune=now` originally did. Instead, it
> generates a cruft pack in the directory specified by expire-to.

Is this less important than "we added --expire-to to gc that is passed down to underlying repack" in the previous paragraph?

Not removing the unreachables too early with "repack -a" is an essential part of the design of this new feature to allow us not to lose the cruft objects, so I was a bit surprised that this was described as a "Note:".

Show 13 quoted lines

> diff --git a/Documentation/git-gc.txt b/Documentation/git-gc.txt
> index 370e22faaeb..b4c0cf02972 100644
> --- a/Documentation/git-gc.txt
> +++ b/Documentation/git-gc.txt
> @@ -69,6 +69,12 @@ be performed as well.
>  	the `--max-cruft-size` option of linkgit:git-repack[1] for
>  	more.
>  
> +--expire-to=<dir>::
> +	When packing unreachable objects into a cruft pack, write a cruft
> +	pack containing pruned objects (if any) to the directory `<dir>`.
> +	See the `--expire-to` option of linkgit:git-repack[1] for
> +	more.

Does "When packing unreachable objects into a cruft pack" mean that this option is only meaningful with "--cruft"? As "--cruft" is on by default, is it an error to pass "--no-cruft" when you use this option?

"for more" -> "for more information" or something?

Show 20 quoted lines

> diff --git a/builtin/gc.c b/builtin/gc.c
> index d52735354c9..8656e1caff0 100644
> --- a/builtin/gc.c
> +++ b/builtin/gc.c
> @@ -136,6 +136,7 @@ struct gc_config {
>  	char *prune_worktrees_expire;
>  	char *repack_filter;
>  	char *repack_filter_to;
> +	char *repack_expire_to;
>  	unsigned long big_pack_threshold;
>  	unsigned long max_delta_cache_size;
>  };
> @@ -432,7 +433,8 @@ static int keep_one_pack(struct string_list_item *item, void *data UNUSED)
>  static void add_repack_all_option(struct gc_config *cfg,
>  				  struct string_list *keep_pack)
>  {
> -	if (cfg->prune_expire && !strcmp(cfg->prune_expire, "now"))
> +	if (cfg->prune_expire && !strcmp(cfg->prune_expire, "now")
> +		&& !(cfg->cruft_packs && cfg->repack_expire_to))
>  		strvec_push(&repack, "-a");

Hmph. When "--expire-to=<there>" is given, we are dropping these unreachable objects right away, but we said "--no-cruft", then we say "repack -a". If we have both "--cruft" and "--expire-to=<there>", then ...

Show 8 quoted lines

>  	else if (cfg->cruft_packs) {
>  		strvec_push(&repack, "--cruft");
> @@ -441,6 +443,8 @@ static void add_repack_all_option(struct gc_config *cfg,
>  		if (cfg->max_cruft_size)
>  			strvec_pushf(&repack, "--max-cruft-size=%lu",
>  				     cfg->max_cruft_size);
> +		if (cfg->repack_expire_to)
> +			strvec_pushf(&repack, "--expire-to=%s", cfg->repack_expire_to);

... we do the usual "repack --cruft --expire-to=<there>" in the next block.

Show 8 quoted lines

> @@ -675,7 +679,6 @@ struct repository *repo UNUSED)
>  	const char *prune_expire_sentinel = "sentinel";
>  	const char *prune_expire_arg = prune_expire_sentinel;
>  	int ret;
> -
>  	struct option builtin_gc_options[] = {
>  		OPT__QUIET(&quiet, N_("suppress progress reporting")),
>  		{ OPTION_STRING, 0, "prune", &prune_expire_arg, N_("date"),

OK.

Show 8 quoted lines

> @@ -694,6 +697,8 @@ struct repository *repo UNUSED)
>  			   PARSE_OPT_NOCOMPLETE),
>  		OPT_BOOL(0, "keep-largest-pack", &keep_largest_pack,
>  			 N_("repack all other packs except the largest pack")),
> +		OPT_STRING(0, "expire-to", &cfg.repack_expire_to, N_("dir"),
> +			   N_("pack prefix to store a pack containing pruned objects")),
>  		OPT_END()
>  	};

OK.

Show 13 quoted lines

> diff --git a/t/t6500-gc.sh b/t/t6500-gc.sh
> index ee074b99b70..d4b0653a9b7 100755
> --- a/t/t6500-gc.sh
> +++ b/t/t6500-gc.sh
> @@ -339,6 +339,12 @@ test_expect_success 'gc.maxCruftSize sets appropriate repack options' '
>  	test_subcommand $cruft_max_size_opts --max-cruft-size=3145728 <trace2.txt
>  '
>  
> +test_expect_success '--expire-to sets appropriate repack options' '
> +	mkdir expired &&
> +	GIT_TRACE2_EVENT=$(pwd)/trace2.txt git -C cruft--max-size gc --cruft --expire-to=./expired/pack &&
> +	test_subcommand $cruft_max_size_opts --expire-to=./expired/pack <trace2.txt
> +'

As "--cruft" is on by default, the command line does not have to have it, but being explicit is good.

Should we also see what happens when "--no-cruft" is given?

Thanks.