From d12d7952f1fa4e511275b6b70c0d25fa0efd50ff Mon Sep 17 00:00:00 2001
From: Yuri Kunde Schlesner <yuriks@yuriks.net>
Date: Sat, 10 May 2014 21:48:46 -0300
Subject: [PATCH] Align pointer in SDL_memset before doing Uint32 loop

Some more recent compilers emit SSE aligned store instructions for the loop,
causing crashes if the destination buffer isn't aligned on a 32-bit boundary.
This would also crash on platforms like ARM that require aligned stores.

This fixes a crash inside SDL_FillRect that happens with the official x64 mingw
build.
---
 src/stdlib/SDL_string.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/src/stdlib/SDL_string.c b/src/stdlib/SDL_string.c
index fb50687cc..a9ca2674f 100644
--- a/src/stdlib/SDL_string.c
+++ b/src/stdlib/SDL_string.c
@@ -263,13 +263,25 @@ SDL_memset(void *dst, int c, size_t len)
 #if defined(HAVE_MEMSET)
     return memset(dst, c, len);
 #else
-    size_t left = (len % 4);
+    size_t left;
     Uint32 *dstp4;
-    Uint8 *dstp1;
+    Uint8 *dstp1 = (Uint8 *) dst;
     Uint32 value4 = (c | (c << 8) | (c << 16) | (c << 24));
     Uint8 value1 = (Uint8) c;
 
-    dstp4 = (Uint32 *) dst;
+    /* The destination pointer needs to be aligned on a 4-byte boundary to
+     * execute a 32-bit set. Set first bytes manually if needed until it is
+     * aligned. */
+    while ((intptr_t)dstp1 & 0x3) {
+        if (len--) {
+            *dstp1++ = value1;
+        } else {
+            return dst;
+        }
+    }
+
+    dstp4 = (Uint32 *) dstp1;
+    left = (len % 4);
     len /= 4;
     while (len--) {
         *dstp4++ = value4;