xdiff: import new 32-bit version of count_masked_bytes()

Import the latest 32-bit implementation of count_masked_bytes() from
Linux (arch/x86/include/asm/word-at-a-time.h).  It's shorter and avoids
overflows and negative numbers.

This fixes test failures on 32-bit, where negative partial results had
been shifted right using the "wrong" method (logical shift right instead
of arithmetic short right).  The compiler is free to chose the method,
so it was only wrong in the sense that it didn't work as intended by us.

Reported-by: Øyvind A. Holm <sunny@sunbase.org>
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/xdiff/xutils.c b/xdiff/xutils.c
index 2021117799ac9c6ede0407ae3d943ec30bd05b73..179bc4fe04c62bb203522915df1d4558feaec7c9 100644 (file)
--- a/xdiff/xutils.c
+++ b/xdiff/xutils.c
@@ -309,19 +309,11 @@ static inline long count_masked_bytes(unsigned long mask)
                long a = (REPEAT_BYTE(0x01) / 0xff + 1);
                return mask * a >> (sizeof(long) * 7);
        } else {
-               /*
-                * Modified Carl Chatfield G+ version for 32-bit *
-                *
-                * (a) gives us
-                *   -1 (0, ff), 0 (ffff) or 1 (ffffff)
-                * (b) gives us
-                *   0 for 0, 1 for (ff ffff ffffff)
-                * (a+b+1) gives us
-                *   correct 0-3 bytemask count result
-                */
-               long a = (mask - 256) >> 23;
-               long b = mask & 1;
-               return a + b + 1;
+               /* Carl Chatfield / Jan Achrenius G+ version for 32-bit */
+               /* (000000 0000ff 00ffff ffffff) -> ( 1 1 2 3 ) */
+               long a = (0x0ff0001 + mask) >> 23;
+               /* Fix the 1 for 00 case */
+               return a & mask;
        }
 }