If a word loaded from the string being compared contains the byte-pair
0x01000 at any position within the word (ie 0xXXXX0100, 0xXX0100XX or
0x0100XXXX) then the syndrome value calculated by the zero-detection
algorithm will contain 0x8080 at the position of these two bytes.
On a little-endian machine this makes no difference to the behaviour of
the algorithm because the interesting bytes (if any) are in the least
significant part of the word; but on big-endian machines the 0x01 byte
is the last byte in the string. We therefore have to handle this case
by testing each of the last bytes explicitly.
2009-03-24 Richard Earnshaw <rearnsha@arm.com>
* libc/machine/arm/strcmp.c (strcmp_unaligned): Correctly
detect the nul-byte in a big-endian string.