reset: fix write race

Similar to the previous commit "commit: fix write race", fix a similar race in
"reset" (except for --soft, which does not lock/write the index) by changing
"read -> lock -> write" to "lock -> read -> write", so "reset" now qualifies
the new user-friendly "waiting for index lock" behavior.

Unlike "commit", "reset" uses repo_read_index(), not repo_read_index_preload(),
so it won't add significant work (e.g. lstat() working tree) while holding the
index lock.

I chose to not add a test to t0032-index-lock-timeout.sh, since every waiting
test case there (undesirably) takes 1 second to run, there are already
representative test cases, and the change looks straightforward.

Signed-off-by: Jun Wu <quark@meta.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

builtin/reset.c

@@ -346,6 +346,7 @@ int cmd_reset(int argc,
 	struct object_id oid;
 	struct pathspec pathspec;
 	int intent_to_add = 0;
+	struct lock_file lock = LOCK_INIT;
 	struct interactive_options interactive_opts = INTERACTIVE_OPTIONS_INIT;
 	const struct option options[] = {
 		OPT__QUIET(&quiet, N_("be quiet, only report errors")),
@@ -477,6 +478,9 @@ int cmd_reset(int argc,
 	if (intent_to_add && reset_type != MIXED)
 		die(_("the option '%s' requires '%s'"), "-N", "--mixed");
 
+	if (reset_type != SOFT)
+		repo_hold_locked_index(the_repository, &lock,
+				       LOCK_DIE_ON_ERROR);
 	if (repo_read_index(the_repository) < 0)
 		die(_("index file corrupt"));
 
@@ -487,9 +491,6 @@ int cmd_reset(int argc,
 		die_if_unmerged_cache(reset_type);
 
 	if (reset_type != SOFT) {
-		struct lock_file lock = LOCK_INIT;
-		repo_hold_locked_index(the_repository, &lock,
-				       LOCK_DIE_ON_ERROR);
 		if (reset_type == MIXED) {
 			int flags = quiet ? REFRESH_QUIET : REFRESH_IN_PORCELAIN;
 			if (read_from_tree(&pathspec, &oid, intent_to_add)) {