From 2bbeddee5d36dfa3fab16ccd0b78dcdcf66c3a99 Mon Sep 17 00:00:00 2001 From: Jeff King Date: Sat, 19 Aug 2023 19:53:42 -0400 Subject: [PATCH] fsck: use enum object_type for fsck_walk callback We switched the function interface for fsck callbacks in a1aad71601 (fsck.h: use "enum object_type" instead of "int", 2021-03-28). However, we accidentally flipped the type back to "int" as part of 0b4e9013f1 (fsck: mark unused parameters in various fsck callbacks, 2023-07-03). The mistake happened because that commit was written before a1aad71601 and rebased forward, and I screwed up while resolving the conflict. Curiously, the compiler does not warn about this mismatch, at least not when using gcc and clang on Linux (nor in any of our CI environments). Based on 28abf260a5 (builtin/fsck.c: don't conflate "int" and "enum" in callback, 2021-06-01), I'd guess that this would cause the AIX xlc compiler to complain. I noticed because clang-18's UBSan now identifies mis-matched function calls at runtime, and does complain of this case when running the test suite. I'm not entirely clear on whether this mismatch is a problem in practice. Compilers are certainly free to make enums smaller than "int" if they don't need the bits, but I suspect that they have to promote back to int for function calls (though I didn't dig in the standard, and I won't be surprised if I'm simply wrong and the real-world impact would depend on the ABI). Regardless, switching it back to enum is obviously the right thing to do here; the switch to "int" was simply a mistake. Signed-off-by: Jeff King Signed-off-by: Junio C Hamano --- builtin/fsck.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/builtin/fsck.c b/builtin/fsck.c index c1d02900268b03..611925905e4fd1 100644 --- a/builtin/fsck.c +++ b/builtin/fsck.c @@ -206,7 +206,7 @@ static int traverse_reachable(void) return !!result; } -static int mark_used(struct object *obj, int type UNUSED, +static int mark_used(struct object *obj, enum object_type type UNUSED, void *data UNUSED, struct fsck_options *options UNUSED) { if (!obj)