Re: Segfault in git when using git logs

From: Junio C Hamano <gitster@pobox.com>
Date: Nov 2, 2020, 18:31 UTC
Message-ID: <xmqqzh3zvce5.fsf@gitster.c.googlers.com>
In-Reply-To: <20201102144321.GA3962443@coredump.intra.peff.net>

Jeff King <peff@peff.net> writes:

Show 23 quoted lines

> That commit causes the line-log to clear the set of pathspecs, but the
> --follow option requires exactly one pathspec (and it even makes sure
> the user gives us one, but that happens before we clear it internally).
> Something like this makes the segfault go away:
>
> diff --git a/line-log.c b/line-log.c
> index 42c5e41f68..f789863928 100644
> --- a/line-log.c
> +++ b/line-log.c
> @@ -847,6 +847,7 @@ static void queue_diffs(struct line_log_data *range,
>  		clear_pathspec(&opt->pathspec);
>  		parse_pathspec_from_ranges(&opt->pathspec, range);
>  	}
> +	opt->flags.follow_renames = 0;
>  	DIFF_QUEUE_CLEAR(&diff_queued_diff);
>  	diff_tree_oid(parent_tree_oid, tree_oid, "", opt);
>  	if (opt->detect_rename && diff_might_be_rename()) {
>
> but I'm not clear on how "--follow" and "-L" are supposed to interact. I
> wouldn't expect --follow to do anything at all with line-log (nor for it
> to be useful to specify pathspecs outside of the -L option). So possibly
> this is restoring the behavior prior to that commit, or possibly it's
> just papering over a breakage. ;)

Another option is to catch it as "these options are mutually exclusive" error early before the control reaches this deep in the codeflow, I would think, but I suspect that some people may habitually add the "--follow" option in a context where it does not make sense, so "--follow is silently ignored when other options that contradict it is in effect at the same time" is OK by me, too.

I do not know if that is the case offhand, but in the ideal world, I would imagine

	git log -L1,5:hello.c -C -C -- hello.c goodbye.c
	git log -L1,5:hello.c -C -C

to notice and show that some of these five lines were copied when or after hello.c was created, and keep following the origin of them in goodbye.c, and

	git log -L1,5:hello.c -C -C

may do the same or find better match outside goodbye.c for the origin of these lines and follow them instead, while

	git log -L1,5:hello.c -- hello.c
	git log -L1,5:hello.c --no-renames

in the same history may say the commit that actually copied these lines from goodbye.c just added them directly to hello.c instead.

And to extend the imagination a bit more,

	git log -M -L1,5:hello.c
	git log -L1,5:hello.c
	git log --follow -L1,5:hello.c

in a different history may notice that hello.c was created wholesale by renaming hola.c and follow the changes to these five lines down the history. As -M is in effect by default these days, the first two would be equivalent, and "--follow" would be a meaningless noiseword in the context of this example where we are interested only in a single path hello.c in the end result, but in the ideal world, meaningless noiseword should become hardless no-op, I would think.

Of course, the above assumes that -L plays well with the -M/-C/--follow options and pathspec. If not, then "they are incompatible" would be the more sensible and easy-to-explain option.

Thanks.