public inbox for git@vger.kernel.org
 help / color / mirror / Atom feed
From: Junio C Hamano <gitster@pobox.com>
To: Adrian Ratiu <adrian.ratiu@collabora.com>
Cc: git@vger.kernel.org, "Jeff King" <peff@peff.net>,
	"Emily Shaffer" <emilyshaffer@google.com>,
	"Patrick Steinhardt" <ps@pks.im>,
	"Josh Steadmon" <steadmon@google.com>,
	"Kristoffer Haugsbakk" <kristofferhaugsbakk@fastmail.com>,
	"brian m . carlson" <sandals@crustytoothpaste.net>,
	"Emily Shaffer" <nasamuffin@google.com>,
	"Ævar Arnfjörð Bjarmason" <avarab@gmail.com>
Subject: Re: [PATCH v3 4/9] hook: allow parallel hook execution
Date: Sun, 15 Mar 2026 13:46:57 -0700	[thread overview]
Message-ID: <xmqqjyvcstvi.fsf@gitster.g> (raw)
In-Reply-To: <20260309133739.294555-5-adrian.ratiu@collabora.com> (Adrian Ratiu's message of "Mon, 9 Mar 2026 15:37:34 +0200")

Adrian Ratiu <adrian.ratiu@collabora.com> writes:

> +hook.<name>.parallel::
> +	Whether the hook `hook.<name>` may run in parallel with other hooks
> +	for the same event. Defaults to `false`. Set to `true` only when the
> +	hook script is safe to run concurrently with other hooks for the same
> +	event. If any hook for an event does not have this set to `true`,
> +	all hooks for that event run sequentially regardless of `hook.jobs`.

This is very conservative and safe default.

> @@ -307,6 +316,7 @@ static void build_hook_config_map(struct repository *r,
>  			entry->command = command ? xstrdup(command) : NULL;
>  			entry->scope = scope;
>  			entry->disabled = is_disabled;
> +			entry->parallel = (int)(uintptr_t)par;

Hmm.  The source "par" is (void *) and the destination .parallel
member is a single bit, so would this

			entry->parallel = !!par;

be the same?  A cast first to uintptr_t, presumably not to lose
bits, and then casting it down to potentially narrower int made me
wonder what else is going on here that is tricky.

> +/* Determine how many jobs to use for hook execution. */
> +static unsigned int get_hook_jobs(struct repository *r,
> +				  struct run_hooks_opt *options,
> +				  struct string_list *hook_list)
> +{
> +	unsigned int jobs;
> +
> +	/*
> +	 * Hooks needing separate output streams must run sequentially. Next
> +	 * commits will add an extension to allow parallelizing these as well.
> +	 */
> +	if (!options->stdout_to_stderr)
> +		return 1;
> +
> +	/* An explicit job count (FORCE_SERIAL jobs=1, or -j from CLI). */
> +	if (options->jobs)
> +		return options->jobs;

This could be risky for hooks that claim they do not want to run
with others at the same time, but the CLI user ought to know what
they are using, so this override is very much appreciated.  After
all, the override may be serializing an overly optimisitic set of
hooks that want to run in parallel to avoid interaction between
them.

> +	/*
> +	 * Use hook.jobs from the already-parsed config cache (in-repo), or
> +	 * fall back to a direct config lookup (out-of-repo).  Default to 1.
> +	 */
> +	if (r && r->gitdir && r->hook_config_cache)
> +		/* Use the already-parsed cache (in-repo) */
> +		jobs = r->hook_config_cache->jobs ? r->hook_config_cache->jobs : 1;
> +	else
> +		/* No cache present (out-of-repo call), use direct cfg lookup */
> +		jobs = repo_config_get_uint(r, "hook.jobs", &jobs) ? 1 : jobs;
> +
> +	/*
> +	 * Cap to serial any configured hook not marked as parallel = true.
> +	 * This enforces the parallel = false default, even for "traditional"
> +	 * hooks from the hookdir which cannot be marked parallel = true.
> +	 */
> +	for (size_t i = 0; jobs > 1 && i < hook_list->nr; i++) {
> +		struct hook *h = hook_list->items[i].util;
> +		if (h->kind == HOOK_CONFIGURED && !h->parallel)
> +			jobs = 1;
> +	}

Losing "jobs > 1 &&" from the termination condition and instead
explicitly "break;" out when we demote jobs to 1 would be easier to
read, even though it would spend two more lines, i.e.,

	for (size_t i = 0; i < hook_list->nr; i++) {
        	struct hook *h = ...;
		if (...) {
			jobs = 1;
			break;
		}
	}


Other than that, very cleanly written.

Thanks.

  reply	other threads:[~2026-03-15 20:47 UTC|newest]

Thread overview: 83+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-02-04 17:33 [PATCH 0/4] Run hooks in parallel Adrian Ratiu
2026-02-04 17:33 ` [PATCH 1/4] config: add a repo_config_get_uint() helper Adrian Ratiu
2026-02-04 17:33 ` [PATCH 2/4] hook: allow parallel hook execution Adrian Ratiu
2026-02-11 12:41   ` Patrick Steinhardt
2026-02-12 12:25     ` Adrian Ratiu
2026-02-04 17:33 ` [PATCH 3/4] hook: introduce extensions.hookStdoutToStderr Adrian Ratiu
2026-02-04 17:33 ` [PATCH 4/4] hook: allow runtime enabling extensions.hookStdoutToStderr Adrian Ratiu
2026-02-12 10:43 ` [PATCH 0/4] Run hooks in parallel Phillip Wood
2026-02-12 14:24   ` Adrian Ratiu
2026-02-13 14:39     ` Phillip Wood
2026-02-13 17:21       ` Adrian Ratiu
2026-02-22  0:28 ` [PATCH v2 00/10] " Adrian Ratiu
2026-02-22  0:28   ` [PATCH v2 01/10] repository: fix repo_init() memleak due to missing _clear() Adrian Ratiu
2026-02-22  0:28   ` [PATCH v2 02/10] config: add a repo_config_get_uint() helper Adrian Ratiu
2026-02-22  0:28   ` [PATCH v2 03/10] hook: refactor hook_config_cache from strmap to named struct Adrian Ratiu
2026-02-22  0:28   ` [PATCH v2 04/10] hook: parse the hook.jobs config Adrian Ratiu
2026-02-22  0:28   ` [PATCH v2 05/10] hook: allow parallel hook execution Adrian Ratiu
2026-02-22  0:29   ` [PATCH v2 06/10] hook: mark non-parallelizable hooks Adrian Ratiu
2026-02-22  0:29   ` [PATCH v2 07/10] hook: add -j/--jobs option to git hook run Adrian Ratiu
2026-02-22  0:29   ` [PATCH v2 08/10] hook: add per-event jobs config Adrian Ratiu
2026-02-22  0:29   ` [PATCH v2 09/10] hook: introduce extensions.hookStdoutToStderr Adrian Ratiu
2026-02-22  0:29   ` [PATCH v2 10/10] hook: allow runtime enabling extensions.hookStdoutToStderr Adrian Ratiu
2026-03-09 13:37 ` [PATCH v3 0/9] Run hooks in parallel Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 1/9] repository: fix repo_init() memleak due to missing _clear() Adrian Ratiu
2026-03-15  4:55     ` Junio C Hamano
2026-03-15  5:05     ` Junio C Hamano
2026-03-09 13:37   ` [PATCH v3 2/9] config: add a repo_config_get_uint() helper Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 3/9] hook: parse the hook.jobs config Adrian Ratiu
2026-03-15 16:13     ` Junio C Hamano
2026-03-09 13:37   ` [PATCH v3 4/9] hook: allow parallel hook execution Adrian Ratiu
2026-03-15 20:46     ` Junio C Hamano [this message]
2026-03-18 18:02       ` Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 5/9] hook: mark non-parallelizable hooks Adrian Ratiu
2026-03-15 20:56     ` Junio C Hamano
2026-03-18 18:40       ` Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 6/9] hook: add -j/--jobs option to git hook run Adrian Ratiu
2026-03-15 21:00     ` Junio C Hamano
2026-03-18 19:00       ` Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 7/9] hook: add per-event jobs config Adrian Ratiu
2026-03-16 18:40     ` Junio C Hamano
2026-03-18 19:21       ` Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 8/9] hook: introduce extensions.hookStdoutToStderr Adrian Ratiu
2026-03-16 18:44     ` Junio C Hamano
2026-03-18 19:50       ` Adrian Ratiu
2026-03-09 13:37   ` [PATCH v3 9/9] hook: allow runtime enabling extensions.hookStdoutToStderr Adrian Ratiu
2026-03-20 13:53 ` [PATCH v4 0/9] Run hooks in parallel Adrian Ratiu
2026-03-20 13:53   ` [PATCH v4 1/9] config: add a repo_config_get_uint() helper Adrian Ratiu
2026-03-20 13:53   ` [PATCH v4 2/9] hook: parse the hook.jobs config Adrian Ratiu
2026-03-24  9:07     ` Patrick Steinhardt
2026-03-24 18:59       ` Adrian Ratiu
2026-03-20 13:53   ` [PATCH v4 3/9] hook: allow parallel hook execution Adrian Ratiu
2026-03-24  9:07     ` Patrick Steinhardt
2026-03-20 13:53   ` [PATCH v4 4/9] hook: allow pre-push parallel execution Adrian Ratiu
2026-03-20 13:53   ` [PATCH v4 5/9] hook: mark non-parallelizable hooks Adrian Ratiu
2026-03-20 13:53   ` [PATCH v4 6/9] hook: add -j/--jobs option to git hook run Adrian Ratiu
2026-03-24  9:07     ` Patrick Steinhardt
2026-03-20 13:53   ` [PATCH v4 7/9] hook: add per-event jobs config Adrian Ratiu
2026-03-24  9:08     ` Patrick Steinhardt
2026-03-20 13:53   ` [PATCH v4 8/9] hook: warn when hook.<friendly-name>.jobs is set Adrian Ratiu
2026-03-24  9:08     ` Patrick Steinhardt
2026-03-20 13:53   ` [PATCH v4 9/9] hook: add hook.<event>.enabled switch Adrian Ratiu
2026-03-24  9:08     ` Patrick Steinhardt
2026-03-25 18:43       ` Adrian Ratiu
2026-03-20 17:24   ` [PATCH v4 0/9] Run hooks in parallel Junio C Hamano
2026-03-23 15:07     ` Adrian Ratiu
2026-03-24  9:07       ` Patrick Steinhardt
2026-03-26 10:18 ` [PATCH v5 00/12] " Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 01/12] repository: fix repo_init() memleak due to missing _clear() Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 02/12] config: add a repo_config_get_uint() helper Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 03/12] hook: parse the hook.jobs config Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 04/12] hook: allow parallel hook execution Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 05/12] hook: allow pre-push parallel execution Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 06/12] hook: mark non-parallelizable hooks Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 07/12] hook: add -j/--jobs option to git hook run Adrian Ratiu
2026-03-27 14:46     ` Patrick Steinhardt
2026-03-26 10:18   ` [PATCH v5 08/12] hook: add per-event jobs config Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 09/12] hook: warn when hook.<friendly-name>.jobs is set Adrian Ratiu
2026-03-27 14:46     ` Patrick Steinhardt
2026-03-26 10:18   ` [PATCH v5 10/12] hook: move is_known_hook() to hook.c for wider use Adrian Ratiu
2026-03-27 14:46     ` Patrick Steinhardt
2026-03-27 15:59       ` Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 11/12] hook: add hook.<event>.enabled switch Adrian Ratiu
2026-03-26 10:18   ` [PATCH v5 12/12] hook: allow hook.jobs=-1 to use all available CPU cores Adrian Ratiu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=xmqqjyvcstvi.fsf@gitster.g \
    --to=gitster@pobox.com \
    --cc=adrian.ratiu@collabora.com \
    --cc=avarab@gmail.com \
    --cc=emilyshaffer@google.com \
    --cc=git@vger.kernel.org \
    --cc=kristofferhaugsbakk@fastmail.com \
    --cc=nasamuffin@google.com \
    --cc=peff@peff.net \
    --cc=ps@pks.im \
    --cc=sandals@crustytoothpaste.net \
    --cc=steadmon@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox