Re: [PATCH v4 3/3] format-patch: teach --cover-from-description option

From: Junio C Hamano <gitster@pobox.com>
Date: Oct 12, 2019, 02:36 UTC
Message-ID: <xmqqd0f2elli.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <9b8c3dcb539054ba483fc34c6ff509e4ca73517c.1570821015.git.liu.denton@gmail.com>

Denton Liu <liu.denton@gmail.com> writes:

Show 32 quoted lines

> +format.coverFromDescription::
> +	The default mode for format-patch to determine which parts of
> +	the cover letter will be populated using the branch's
> +	description. See the `--cover-from-description` option in
> +	linkgit:git-format-patch[1].
> +
>  format.signature::
>  	The default for format-patch is to output a signature containing
>  	the Git version number. Use this variable to change that default.
> diff --git a/Documentation/git-format-patch.txt b/Documentation/git-format-patch.txt
> index 0ac56f4b70..86114e4c22 100644
> --- a/Documentation/git-format-patch.txt
> +++ b/Documentation/git-format-patch.txt
> @@ -19,6 +19,7 @@ SYNOPSIS
>  		   [--start-number <n>] [--numbered-files]
>  		   [--in-reply-to=<message id>] [--suffix=.<sfx>]
>  		   [--ignore-if-in-upstream]
> +		   [--cover-from-description=<mode>]
>  		   [--rfc] [--subject-prefix=<subject prefix>]
>  		   [(--reroll-count|-v) <n>]
>  		   [--to=<email>] [--cc=<email>]
> @@ -171,6 +172,26 @@ will want to ensure that threading is disabled for `git send-email`.
>  	patches being generated, and any patch that matches is
>  	ignored.
>  
> +--cover-from-description=<mode>::
> +	Controls which parts of the cover letter will be automatically
> +	populated using the branch's description.
> ++
> +If `<mode>` is `message` or `default`, the cover letter subject will be
> +populated with placeholder text. The body of the cover letter will be
> +populated with the branch's description.

I understand that this is what we do now, so those who want to live in the past can set the configuration variable to 'message'.

> +If `<mode>` is `subject`, the beginning of the branch description (up to
> +the first blank line) will populate the cover letter subject. The
> +remainder of the description will populate the body of the cover
> +letter.

s/the beginning of .*blank line)/the first paragraph of the branch description/ may be shorter, but the above is OK, too.

When description is prepared appropriately, this mode would fill both subject and body, which sounds sensible.

> +If `<mode>` is `auto`, if the beginning of the branch description (up to
> +the first line) is greater than 100 characters then the mode will be
> +`message`, otherwise `subject` will be used.

I understand that this is a more clever and safer variant of 'subject'. Do you want to say 100 characters or 100 bytes?

> +If `<mode>` is `none`, both the cover letter subject and body will be
> +populated with placeholder text.

OK, this is done for completeness? I wonder who finds it useful to set it to 'none' *AND* set the branch description. Not a rhetorical question that suggests removing this choice, but purely soliciting opinions from others.

It is unclear (other than the mode word being 'default' for one of the choices) what the new default mode of operation is after the patch is applied among the four presented mode. "This is the default when no configuration nor command line option specifies the desired mode" or something may want to be added to one of these paragraphs.

Show 26 quoted lines

> @@ -1061,13 +1076,16 @@ static void make_cover_letter(struct rev_info *rev, int use_stdout,
>  			      struct commit *origin,
>  			      int nr, struct commit **list,
>  			      const char *branch_name,
> +			      enum cover_from_description cover_from_description_mode,
>  			      int quiet)
>  {
>  	const char *committer;
> -	const char *body = "*** SUBJECT HERE ***\n\n*** BLURB HERE ***\n";
> -	const char *msg;
> +	const char *subject = "*** SUBJECT HERE ***";
> +	const char *body = "*** BLURB HERE ***";
>  	struct shortlog log;
>  	struct strbuf sb = STRBUF_INIT;
> +	struct strbuf description_sb = STRBUF_INIT;
> +	struct strbuf subject_sb = STRBUF_INIT;
>  	int i;
>  	const char *encoding = "UTF-8";
>  	int need_8bit_cte = 0;
> @@ -1095,17 +1113,34 @@ static void make_cover_letter(struct rev_info *rev, int use_stdout,
>  	if (!branch_name)
>  		branch_name = find_branch_name(rev);
>  
> -	msg = body;
> +	if (branch_name && *branch_name)
> +		read_branch_desc(&description_sb, branch_name);

It may not matter in practice but strictly speaking there is no need to read the description if we know that the mode is NONE. Removing the support for the NONE mode may be an easier fix than adding "&& mode != NONE" to the if () condition guarding this call---I dunno.

Show 12 quoted lines

> +	if (cover_from_description_mode != COVER_FROM_NONE && description_sb.len) {
> +		if (cover_from_description_mode == COVER_FROM_SUBJECT ||
> +				cover_from_description_mode == COVER_FROM_AUTO)
> +			body = format_subject(&subject_sb, description_sb.buf, " ");
> +
> +		if (cover_from_description_mode == COVER_FROM_MESSAGE ||
> +				(cover_from_description_mode == COVER_FROM_AUTO &&
> +				 subject_sb.len > COVER_FROM_AUTO_MAX_SUBJECT_LEN))
> +			body = description_sb.buf;
> +		else
> +			subject = subject_sb.buf;
> +	}

I wonder if it make the end result cleaner and easier to follow to replace all of the above with a single line:

	cover_from_desc(&subject, &body, branch_name, desc_mode);

in this caller, and move the logic (and a handful of strbuf used as its implementation detail) into the helper function, including the choice of the default "*** SOMETHING HERE ***", etc., and make the helper *always* return allocated piece of memory in subject and body so that this caller can unconditionally free them.

Thanks.