Re: [RFC 2/2] exclude-promisor-objects: declare when option is allowed

From: Jeff King <peff@peff.net>
To: Matthew DeVore <matvore@google.com>
Cc: git@vger.kernel.org, gitster@pobox.com, pclouds@gmail.com,
	jonathantanmy@google.com, jeffhost@microsoft.com,
	matvore@comcast.net
Subject: Re: [RFC 2/2] exclude-promisor-objects: declare when option is allowed
Date: Wed, 21 Nov 2018 11:40:19 -0500	[thread overview]
Message-ID: <20181121164019.GA24621@sigill.intra.peff.net> (raw)
In-Reply-To: <931421945c040ba4518d91f7af9f386d0136bd2f.1540256910.git.matvore@google.com>

On Mon, Oct 22, 2018 at 06:13:42PM -0700, Matthew DeVore wrote:

> diff --git a/builtin/prune.c b/builtin/prune.c
> index 41230f8215..11284d0bf3 100644
> --- a/builtin/prune.c
> +++ b/builtin/prune.c
> @@ -120,6 +120,7 @@ int cmd_prune(int argc, const char **argv, const char *prefix)
>  	save_commit_buffer = 0;
>  	read_replace_refs = 0;
>  	ref_paranoia = 1;
> +	revs.allow_exclude_promisor_objects_opt = 1;
>  	repo_init_revisions(the_repository, &revs, prefix);
>  
>  	argc = parse_options(argc, argv, prefix, options, prune_usage, 0);

I think this line is in the wrong place. The very first thing
repo_init_revisions() will do is memset() the revs struct to all-zeroes,
so it cannot possibly be doing anything.

Normally it would need to go after init_revisions() but before
setup_revisions(), but we don't seem to call the latter at all in
builtin/prune.c. Which makes sense, because you cannot pass options to
influence the reachability traversal. So I don't think we need to care
about this flag at all here.

Speaking of which, would this flag work better as a field in
setup_revision_opt, which is passed to setup_revisions()? The intent
seem to be to influence how we parse command-line arguments, and that's
where other similar flags are (e.g., assume_dashdash).

-Peff