Oh I see you mean now: you thought I should check
host_words_big_endian against BITS_BIG_ENDIAN. That was
deliberate; I don't want to assume BITS_BIG_ENDIAN from
HOST_WORDS_BIG_ENDIAN. It'd just feel very uncomfortable and
goes against my experience; no hard facts, sorry. Just look
what happen when I assumed that I should check
HOST_FLOAT_WORDS_BIG_ENDIAN against FLOAT_WORDS_BIG_ENDIAN here;
arm-elf broke.
Is the patch ok to commit, trunk and 3.2 branch?
It's OK for the trunk, but I have no authority over the branch.