Shaoxuan Yuan wrote:
> Originally, moving a <source> directory which is not on-disk due
> to its existence outside of sparse-checkout cone, "giv mv" command
> errors out with "bad source".
>
> Add a helper check_dir_in_index() function to see if a directory
> name exists in the index. Also add a SKIP_WORKTREE_DIR bit to mark
> such directories.
>
> Change the checking logic, so that such <source> directory makes
> "giv mv" command warns with "advise_on_updating_sparse_paths()"
> instead of "bad source"; also user now can supply a "--sparse" flag so
> this operation can be carried out successfully.
>
> Signed-off-by: Shaoxuan Yuan <shaoxuan.yuan02@xxxxxxxxx>
> ---
> builtin/mv.c | 49 +++++++++++++++++++++++++++++++----
> t/t7002-mv-sparse-checkout.sh | 4 +--
> 2 files changed, 46 insertions(+), 7 deletions(-)
>
> diff --git a/builtin/mv.c b/builtin/mv.c
> index 7ce7992d6c..cb3441c7cb 100644
> --- a/builtin/mv.c
> +++ b/builtin/mv.c
> @@ -123,6 +123,37 @@ static int index_range_of_same_dir(const char *src, int length,
> return last - first;
> }
>
> +/*
> + * Check if an out-of-cone directory should be in the index. Imagine this case
> + * that all the files under a directory are marked with 'CE_SKIP_WORKTREE' bit
> + * and thus the directory is sparsified.> + *
> + * Return 0 if such directory exist (i.e. with any of its contained files not
> + * marked with CE_SKIP_WORKTREE, the directory would be present in working tree).
> + * Return 1 otherwise.
> + */
This explanation is helpful in clarifying that you don't mean *sparse
directories* (that is, directory entries in a sparse index), you mean
directories whose contents are all sparse. It's a tricky distinction, but
you handled it nicely here.
> +static int check_dir_in_index(const char *name, int namelen)
> +{
> + int ret = 1;
> + const char *with_slash = add_slash(name);
> + int length = namelen + 1;
> +
> + int pos = cache_name_pos(with_slash, length);
> + const struct cache_entry *ce;
> +
> + if (pos < 0) {
> + pos = -pos - 1;
> + if (pos >= the_index.cache_nr)
> + return ret;
> + ce = active_cache[pos];
> + if (strncmp(with_slash, ce->name, length))
> + return ret;
> + if (ce_skip_worktree(ce))
> + return ret = 0;
> + }
> + return ret;
The way 'ret' is handled here is a bit difficult to follow. Would you be
opposed to returning hardcoded '0' or '1', rather than changing the value of
'ret' throughout? Something like:
static int check_dir_in_index(const char *name, int namelen)
{
int pos, length = namelen + 1;
const struct cache_entry *ce;
const char *with_slash = add_slash(name);
pos = cache_name_pos(with_slash, length);
if (pos < 0) {
pos = -pos - 1;
if (pos >= the_index.cache_nr)
return 1;
ce = active_cache[pos];
if (strncmp(with_slash, ce->name, length))
return 1;
if (ce_skip_worktree(ce))
return 0;
}
return 1;
}
> +}
> +
> int cmd_mv(int argc, const char **argv, const char *prefix)
> {
> int i, flags, gitmodules_modified = 0;
> @@ -184,7 +215,7 @@ int cmd_mv(int argc, const char **argv, const char *prefix)
> /* Checking */
> for (i = 0; i < argc; i++) {
> const char *src = source[i], *dst = destination[i];
> - int length, src_is_dir;
> + int length;
> const char *bad = NULL;
> int skip_sparse = 0;
>
> @@ -198,12 +229,17 @@ int cmd_mv(int argc, const char **argv, const char *prefix)
>
> pos = cache_name_pos(src, length);
> if (pos < 0) {
> + const char *src_w_slash = add_slash(src);
> + if (!check_dir_in_index(src, length) &&
> + !path_in_sparse_checkout(src_w_slash, &the_index)) {
In checks like these, the less "expensive" one should come first (so that if
it returns 'false', we completely skip the more expensive one). Since
'check_dir_in_index()' requires binary searching the index, it's likely to
be more expensive than 'path_in_sparse_checkout()', so the condition order
should be flipped:
if (!path_in_sparse_checkout(src_w_slash, &the_index) &&
!check_dir_in_index(src, length)) {
Also nit: alignment (more details on why/how in my last message [1]).
[1] https://lore.kernel.org/git/01b39c63-5652-4293-0424-ff99b6f9f7d2@xxxxxxxxxx/
> + modes[i] |= SKIP_WORKTREE_DIR;
> + goto dir_check;
> + }
> /* only error if existence is expected. */
> if (!(modes[i] & SPARSE))
> bad = _("bad source");
> goto act_on_entry;
> }
> -
> ce = active_cache[pos];
> if (!ce_skip_worktree(ce)) {
> bad = _("bad source");
> @@ -230,14 +266,17 @@ int cmd_mv(int argc, const char **argv, const char *prefix)
> bad = _("can not move directory into itself");
> goto act_on_entry;
> }
> - if ((src_is_dir = S_ISDIR(st.st_mode))
> + if (S_ISDIR(st.st_mode)
> && lstat(dst, &st) == 0) {
> bad = _("cannot move directory over file");
> goto act_on_entry;
> }
> - if (src_is_dir) {
> +
> +dir_check:
> + if (S_ISDIR(st.st_mode)) {
> int j, dst_len, n;
> - int first = cache_name_pos(src, length), last;
> + int first, last;
> + first = cache_name_pos(src, length);
Super-nit: why did this line change? It looks like it just rearranges the
lines for no functional purpose.
>
> if (first >= 0) {
> prepare_move_submodule(src, first,
> diff --git a/t/t7002-mv-sparse-checkout.sh b/t/t7002-mv-sparse-checkout.sh
> index 5b61fbad5f..30e13b9979 100755
> --- a/t/t7002-mv-sparse-checkout.sh
> +++ b/t/t7002-mv-sparse-checkout.sh
> @@ -219,7 +219,7 @@ test_expect_success 'refuse to move file to non-skip-worktree sparse path' '
> test_cmp expect stderr
> '
>
> -test_expect_failure 'refuse to move out-of-cone directory without --sparse' '
> +test_expect_success 'refuse to move out-of-cone directory without --sparse' '
> test_when_finished "cleanup_sparse_checkout" &&
> setup_sparse_checkout &&
>
> @@ -230,7 +230,7 @@ test_expect_failure 'refuse to move out-of-cone directory without --sparse' '
> test_cmp expect stderr
> '
>
> -test_expect_failure 'can move out-of-cone directory with --sparse' '
> +test_expect_success 'can move out-of-cone directory with --sparse' '
> test_when_finished "cleanup_sparse_checkout" &&
> setup_sparse_checkout &&
>
