Work around gcc/ld alignment bug on win32

Fixes problems due to misalignment of static zero arrays (win32 ld can't align .bss properly).

configure

@@ -368,7 +368,7 @@ case $host_cpu in
     x86_64)
         ARCH="X86_64"
         AS="yasm"
-        if [ "$SYS" = MACOSX ];then
+        if [ "$SYS" = MACOSX ]; then
             ASFLAGS="$ASFLAGS -f macho64 -m amd64 -DPIC -DPREFIX"
             if cc_check '' "-arch x86_64"; then
                 CFLAGS="$CFLAGS -arch x86_64"
@@ -718,6 +718,11 @@ if cc_check '' -fno-tree-vectorize ; then
     CFLAGS="$CFLAGS -fno-tree-vectorize"
 fi
 
+if [ $SYS = MINGW -a $ARCH = X86 ] ; then
+    # workaround gcc/ld bug with alignment of static variables/arrays that are initialized to zero
+    cc_check '' -fno-zero-initialized-in-bss && CFLAGS="$CFLAGS -fno-zero-initialized-in-bss"
+fi
+
 if cc_check "stdio.h" "" "fseeko(stdin,0,0);" ; then
     define fseek fseeko
     define ftell ftello

encoder/me.c

@@ -589,8 +589,6 @@ void x264_me_search_ref( x264_t *h, x264_me_t *m, int16_t (*mvc)[2], int i_mvc,
             /* successive elimination by comparing DC before a full SAD,
              * because sum(abs(diff)) >= abs(diff(sum)). */
             uint16_t *sums_base = m->integral;
-            /* due to a GCC bug on some platforms (win32?), zero[] may not actually be aligned.
-             * this is not a problem because it is not used for any SSE instructions. */
             ALIGNED_16( static pixel zero[8*FENC_STRIDE] );
             ALIGNED_ARRAY_16( int, enc_dc,[4] );
             int sad_size = i_pixel <= PIXEL_8x8 ? PIXEL_8x8 : PIXEL_4x4;

encoder/rdo.c

@@ -119,7 +119,7 @@ static ALWAYS_INLINE int cached_satd( x264_t *h, int size, int x, int y )
 
 static inline int ssd_plane( x264_t *h, int size, int p, int x, int y )
 {
-    ALIGNED_16(static pixel zero[16]);
+    ALIGNED_16( static pixel zero[16] );
     int satd = 0;
     pixel *fdec = h->mb.pic.p_fdec[p] + x + y*FDEC_STRIDE;
     pixel *fenc = h->mb.pic.p_fenc[p] + x + y*FENC_STRIDE;

encoder/slicetype.c

@@ -191,7 +191,7 @@ static NOINLINE unsigned int x264_weight_cost_chroma( x264_t *h, x264_frame_t *f
     pixel *src = ref + i_offset;
     ALIGNED_ARRAY_16( pixel, buf, [8*8] );
     int pixoff = 0;
-    ALIGNED_16( static pixel flat[9] ) = {0,0,0,0,0,0,0,0,1}; //hack for win32
+    ALIGNED_16( static pixel flat[8] );
     if( w )
     {
         for( int y = 0; y < i_lines; y += 8, pixoff = y*i_stride )