Re: [PATCH] Also read SQUASH_MSG if a conflict on a merge squash occurred

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 8, 2016, 18:32 UTC
Message-ID: <xmqq60wwlt0s.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <56DE5272.2080009@cs-ware.de>

Sven Strickroth <sven@cs-ware.de> writes:

> Subject: Re: [PATCH] Also read SQUASH_MSG if a conflict on a merge squash occurred

A reader sees this line in the output of "git shortlog --no-merges"; does it sufficiently tell her which Git subcommand is affected by this change, if this is a bugfix or a new feature, i.e. enough for her to decide how important the change is?

We often prefix our log message with the name of the area followed by a colon and describe the purpose of the change, not the means how the objective is achieved, e.g.

    Subject: [PATCH] commit: do not lose SQUASH_MSG contents

    When concluding a conflicted "git merge --squash", the command
    failed to read SQUASH_MSG that was prepared by "git merge", and
    showed only the "# Conflicts:" list of conflicted paths.

Show 9 quoted lines

> diff --git a/builtin/commit.c b/builtin/commit.c
> index d054f84..0405d68 100644
> --- a/builtin/commit.c
> +++ b/builtin/commit.c
> @@ -729,6 +729,12 @@ static int prepare_to_commit(const char *index_file, const char *prefix,
>  		if (strbuf_read_file(&sb, git_path_merge_msg(), 0) < 0)
>  			die_errno(_("could not read MERGE_MSG"));
>  		hook_arg1 = "merge";
> +		/* append SQUASH_MSG here if it exists and a merge --squash was originally performed */

	/*
         * Our multi-line comment reads more like
         * this.  That is, the first slash-asterisk is on its
         * own line, so is the last asterisk-slash.
         */

Show 8 quoted lines

> +		if (!stat(git_path_squash_msg(), &statbuf)) {
> +			if (strbuf_read_file(&sb, git_path_squash_msg(), 0) < 0)
> +				die_errno(_("could not read SQUASH_MSG"));
> +			hook_arg1 = "squash";
> +		}
>  	} else if (!stat(git_path_squash_msg(), &statbuf)) {
>  		if (strbuf_read_file(&sb, git_path_squash_msg(), 0) < 0)
>  			die_errno(_("could not read SQUASH_MSG"));

This reads MERGE_MSG first and then SQUASH_MSG; is that what we really want? When you are resolving a conflicted rebase, you would see the original log message and then conflicts section. What is in the SQUASH_MSG is the moral equivalent of the "original log message" but in a less summarized form, so I suspect that the list of conflicts should come to end.

The duplicated code to read the same file bothers me somewhat.

I wondered if it makes the result easier to follow (and easier to update) if this part of the code is restructured like this:

	if (file_exists(git_path_merge_msg()) ||
            file_exists(git_path_squash_msg())) {
	    if (file_exists(git_path_squash_msg())) {
		read SQUASH_MSG;
	    }
            if (file_exists(git_path_merge_msg()))
            	read MERGE_MSG;
	    }
            hook_arg1 = "merge";
	}

but I am not sure if that structure is better.

Thanks.