Re: [PATCH] rev-list --max-age, --max-count: support --boundary

[Junio C Hamano <junkio@cox.net>]

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> Now, when saying --max-age=<timestamp>, or --max-count=<n>, together
> with --boundary, rev-list prints the boundary commits, i.e. the
> commits which are _just_ not shown without --boundary, i.e. their
> children are, but they aren't.
>
> Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
> ---
>
> 	Of course, this is meant to speed up an otherwise expensive
> 	operation in git-bundle.
>
> 	However, this touches a very sensitive point in core git:
> 	the revision walking machinery. It passes all tests, but
> 	some eyeballing is much appreciated.

I think this looks fine.

Initially I thought setting BOUNDARY_SHOW unconditionally would
be wrong (the "limited" case takes pains to avoid showing
redundant boundary commits), and that was exactly why I said it
would be a bit more involved change, but I do not think of a
topology that needs redundancy removal outside of "limited" case.

> @@ -1336,7 +1342,17 @@ struct commit *get_revision(struct rev_info *revs)
>  	case -1:
>  		break;
>  	case 0:
> -		return NULL;
> +		if (revs->boundary) {
> +			struct commit_list *list = revs->commits;
> +			while (list) {
> +				list->item->object.flags |=
> +					BOUNDARY_SHOW | BOUNDARY;
> +				list = list->next;
> +			}
> +			revs->max_count = -1;
> +			revs->limited = 1;
> +		} else
> +			return NULL;

The resetting of max_count and limited probably needs some
commenting.