Re: [PATCH] Allow git-update-index work on subprojects

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

Alex Riesen <raa.lkml@gmail.com> writes:

> Also, make "git commit -a" work with modifications of subproject HEADs.
>
> ---
>
> This one works with update-index --remove (which is what git-commit -a
> uses). It is ugly. I tried to keep the "F -> D/F" behaviour of
> update-index. Still have to check if "F -> Subproject" works.
>
>  builtin-update-index.c |   45 +++++++++++++++++++++++++--------------------
>  1 files changed, 25 insertions(+), 20 deletions(-)
>
> diff --git a/builtin-update-index.c b/builtin-update-index.c
> index eba756d..d075d50 100644
> --- a/builtin-update-index.c
> +++ b/builtin-update-index.c
> @@ -62,7 +62,7 @@ static int mark_valid(const char *path)
>  
>  static int process_file(const char *path)
>  {
> -	int size, namelen, option, status;
> +	int size, namelen = -1, option, status;
>  	struct cache_entry *ce;
>  	struct stat st;
>  
> @@ -73,7 +73,7 @@ static int process_file(const char *path)
>  	 */
>  	cache_tree_invalidate_path(active_cache_tree, path);
>  
> +	if (!status && S_ISDIR(st.st_mode)) {
>  		/* When we used to have "path" and now we want to add
>  		 * "path/file", we need a way to remove "path" before
>  		 * being able to add "path/file".  However,
> @@ -82,27 +82,32 @@ static int process_file(const char *path)
>  		 * friendly, especially since we can do the opposite
>  		 * case just fine without --force-remove.
>  		 */
> +		namelen = strlen(path);
> +		int pos = cache_name_pos(path, namelen);
> +		if (0 <= pos && S_ISREG(ntohl(active_cache[pos]->ce_mode)) &&
> +		    allow_remove) {
> +			if (remove_file_from_cache(path))
> +				return error("%s: cannot remove from the index", path);
> +			else
> +				return 0;
> +		}
> +	}

If I used to have a symlink S and now the filesystem has a file
S/F which I am running "update-index --add --remove" on, what
happens?

If I have a subproject at path P, and mistakenly try to add path
P/F with "update-index --add --remove P/F", it should be
refused, shouldn't it?