[FFmpeg-devel] [PATCH] avcodec/proresenc_anatoliy: Fix invalid left shift of negative number

[Andreas Rheinhardt]

This fixes ticket #7997 as well as the vsynth*-prores_# FATE-tests
(where * ranges over { 1, 2, 3, _lena } and # over { , _int, _444,
_444_int }).

(Given that prev_dc is in the range -0xC000..0x3FFF, no overflow can
happen upon multiplication with 2.)

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at gmail.com>
---
Actually, #7997 ran into the same issue as #7979 and so it is no longer
reproducible with the given commandline since aef24efb (which fixed
#7979). But it is still reproducible with e.g. the null muxer.

 libavcodec/proresenc_anatoliy.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavcodec/proresenc_anatoliy.c b/libavcodec/proresenc_anatoliy.c
index 0fc79fc1de..1fcb0ae913 100644
--- a/libavcodec/proresenc_anatoliy.c
+++ b/libavcodec/proresenc_anatoliy.c
@@ -224,7 +224,7 @@ static void encode_codeword(PutBitContext *pb, int val, int codebook)
 }
 
 #define QSCALE(qmat,ind,val) ((val) / ((qmat)[ind]))
-#define TO_GOLOMB(val) (((val) << 1) ^ ((val) >> 31))
+#define TO_GOLOMB(val) (((val) * 2) ^ ((val) >> 31))
 #define DIFF_SIGN(val, sign) (((val) >> 31) ^ (sign))
 #define IS_NEGATIVE(val) ((((val) >> 31) ^ -1) + 1)
 #define TO_GOLOMB2(val,sign) ((val)==0 ? 0 : ((val) << 1) + (sign))
-- 
2.20.1