Stefan Beller <sbeller@xxxxxxxxxx> writes:
> + if (!strcmp(arg, "--show-superproject-working-tree")) {
> + const char *superproject = get_superproject_working_tree();
> + if (superproject)
> + puts(superproject);
> + continue;
> + }
Returning the exact string of the path from the API function is
absolutely the right thing. I however have to wonder if rev-parse
need to do the c-quoting unless it is told to show pathnames in its
output without quoting (perhaps with "-z"). Paths from "rev-parse"
(like "--git-dir", "--show-toplevel", etc.) already are excempt from
the usual quoting rules, so doing puts() and nothing else is fine to
be consistent with the existing practice, but in the longer term, I
am sure we would need to revisit so that scripts can handle paths
with funny characters sensibly, but that would be a different topic
if existing ones like "--git-dir" are already unsafe.
> if (!strcmp(arg, "--show-prefix")) {
> if (prefix)
> puts(prefix);
> diff --git a/submodule.c b/submodule.c
> index 3b98766a6b..06473d3646 100644
> --- a/submodule.c
> +++ b/submodule.c
> @@ -1514,3 +1514,86 @@ void absorb_git_dir_into_superproject(const char *prefix,
> strbuf_release(&sb);
> }
> }
> +
> +const char *get_superproject_working_tree(void)
> +{
> +...
> + argv_array_pushl(&cp.args, "--literal-pathspecs", "-C", "..",
> + "ls-files", "--stage", "--full-name", "--", subpath, NULL);
> + strbuf_reset(&sb);
> +...
> + if (starts_with(sb.buf, "160000")) {
> + int super_sub_len;
> + int cwd_len = strlen(cwd);
> + char *super_sub, *super_wt;
> +
> + /*
> + * There is a superproject having this repo as a submodule.
> + * The format is <mode> SP <hash> SP <stage> TAB <full name> LF,
> + * First remove LF, then skip up to \t.
> + */
Looks more or less right but invoke "ls-files -z" and reading the \0
delimited output would be easier; otherwise you would have to worry
about c-unquoting the pathname when the submodule is bound at a path
with funny character (like a double-quote) in it.