homepage

Got a test case that demonstrates a failure? Looks like it works to me...
$ uname -ip
sparc SUNW,Sun-Fire-280R
$ python -c 'import sys; print sys.byteorder'
big
$ python -c 'import sha; print sha.new(open("test", "rb").read()).hexdigest()'
851faf3199d27200abf2750c14ae6451696216a9
$ sha1sum -b test
851faf3199d27200abf2750c14ae6451696216a9 *test
# uname -ip
AMD Sempron(tm) Processor 2800+ AuthenticAMD
# python -c 'import sys; print sys.byteorder'
little
# python -c 'import sha; print sha.new(open("test", "rb").read()).hexdigest()'
851faf3199d27200abf2750c14ae6451696216a9
# sha1sum -b /tmp/test
851faf3199d27200abf2750c14ae6451696216a9 *test
I think your code analysis is wrong. Perhaps you missed the call to longReverse(), which does endianness byte-swapping, at the beginning of the sha_transform() that specifically is commented: "When run on a little-endian CPU we need to perform byte reversal on an array of longwords."