From: Paul Floyd <pjfloyd@wanadoo.fr>
Date: Sat, 6 Sep 2025 11:52:47 +0000 (+0200)
Subject: FreeBSD memalign wrapper: size of 0 now allowed
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=3c7a8f435f7a1f5cc64c23c5f03348e0ce29a98e;p=thirdparty%2Fvalgrind.git

FreeBSD memalign wrapper: size of 0 now allowed

memalign, the random allocator function amongst allocators
---

diff --git a/coregrind/m_replacemalloc/vg_replace_malloc.c b/coregrind/m_replacemalloc/vg_replace_malloc.c
index 9e0591abf..96edc5bfe 100644
--- a/coregrind/m_replacemalloc/vg_replace_malloc.c
+++ b/coregrind/m_replacemalloc/vg_replace_malloc.c
@@ -1840,7 +1840,10 @@ extern int * __error(void) __attribute__((weak));
   *
   * FreeBSD, undocumented,  just calls aligned_alloc
   * with size rounded up to a multiple
-  * of aligment
+  * of aligment (unless the alignment is 0 in which case
+  * it just calls malloc [prior to Feb 2023 this wasn't
+  * handled correctly resulting in a division-by-zero crash
+  * in the size roundup code])
   *
   * jemalloc mininum alignment is 1, must be a power of 2
   * it looks like excessively large alignment causes ENOMEM
@@ -1924,7 +1927,7 @@ extern int * __error(void) __attribute__((weak));
 #define VG_MEMALIGN_NO_ALIGN_ZERO 0
 #endif
 
-#if defined(MUSL_LIBC)
+#if defined(MUSL_LIBC) || defined(VGO_freebsd)
 #define VG_MEMALIGN_NO_SIZE_ZERO 0
 #else
 #define VG_MEMALIGN_NO_SIZE_ZERO 1