On Sun, Mar 25, 2018 at 9:43 AM, Thomas Gummerer [off-list ref] wrote:

On 03/24, Joel Teichroeb wrote:

quoted

---

Missing sign-off?  I saw it's missing in the other patches as well.

Thanks! I always forget to add a sign-off.

quoted

[...]
+
+     if (info->has_u) {
+             struct child_process cp = CHILD_PROCESS_INIT;
+             struct child_process cp2 = CHILD_PROCESS_INIT;
+             int res;
+
+             cp.git_cmd = 1;
+             argv_array_push(&cp.args, "read-tree");
+             argv_array_push(&cp.args, sha1_to_hex(info->u_tree.hash));
+             argv_array_pushf(&cp.env_array, "GIT_INDEX_FILE=%s", stash_index_path);
+
+             cp2.git_cmd = 1;
+             argv_array_pushl(&cp2.args, "checkout-index", "--all", NULL);
+             argv_array_pushf(&cp2.env_array, "GIT_INDEX_FILE=%s", stash_index_path);
+
+             res = run_command(&cp) || run_command(&cp2);
+             remove_path(stash_index_path);
+             if (res)
+                     return error(_("Could not restore untracked files from stash"));

A minor change in behaviour here is that we are removing the temporary
index file unconditionally here, while we would previously only remove
it if both 'read-tree' and 'checkout-index' would succeed.

I don't think that's a bad thing, we probably don't want users to try
and use that index file in any way, and I doubt that's part of anyones
workflow, so I think cleaning it up makes sense.

I'm not sure about that. The shell script has a trap near the start in
order to clean up the temp index, unless I'm understanding the shell
script incorrectly.