[PATCH] name-hash.c: always initialize dir_next pointer

Johannes Sixt· Nov 1, 2011, 22:21 UTC · lore

Test t2021-checkout-overwrite.sh reveals a segfault in 'git add' on a case-insensitive file system when git is compiled with XMALLOC_POISON defined. The reason is that 2548183b (fix phantom untracked files when core.ignorecase is set) added a new member dir_next to struct cache_entry, but forgot to initialize it in all cases.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
---
 You can also insert git config core.ignorecase true before the first
 test script in t2021 to see the segfault.

 I actually found the crash with an MSVC debug build, which has something
 like XMALLOC_POISON built-in.

 name-hash.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

Show changes to name-hash.c +1 −1

diff --git a/name-hash.c b/name-hash.c
index 225dd76..d8d25c2 100644
--- a/name-hash.c
+++ b/name-hash.c
@@ -74,7 +74,7 @@ static void hash_index_entry(struct index_state *istate, struct cache_entry *ce)
 	if (ce->ce_flags & CE_HASHED)
 		return;
 	ce->ce_flags |= CE_HASHED;
-	ce->next = NULL;
+	ce->next = ce->dir_next = NULL;
 	hash = hash_name(ce->name, ce_namelen(ce));
 	pos = insert_hash(hash, ce, &istate->name_hash);
 	if (pos) {

-- 
1.7.7.1.586.ga0958b

Junio C Hamano· Nov 1, 2011, 22:31 UTC · re: Johannes Sixt · lore

Re: [PATCH] name-hash.c: always initialize dir_next pointer

Thanks.

Jeff King· Nov 1, 2011, 22:39 UTC · re: Johannes Sixt · lore

Re: [PATCH] name-hash.c: always initialize dir_next pointer

On Tue, Nov 01, 2011 at 11:21:06PM +0100, Johannes Sixt wrote:

Show 10 quoted lines

> Test t2021-checkout-overwrite.sh reveals a segfault in 'git add' on a
> case-insensitive file system when git is compiled with XMALLOC_POISON
> defined. The reason is that 2548183b (fix phantom untracked files when
> core.ignorecase is set) added a new member dir_next to struct cache_entry,
> but forgot to initialize it in all cases.
> 
> Signed-off-by: Johannes Sixt <j6t@kdbg.org>
> ---
>  You can also insert git config core.ignorecase true before the first
>  test script in t2021 to see the segfault.

Yikes, thanks.

-Peff

patchname-hash.c: always initialize dir_next pointer