------- Comment #13 from roger at eyesopen dot com 2006-11-10 19:41 ------- Grr. Yep, Michael's BITS_BIG_ENDIAN test looks to be the correct thing. The effect of BITS_BIG_ENDIAN on sign_extract and zero_extract is documented in rtl.texi. Is anyone bootstrapping and regression testing this change or shall I do the honors. In fact, I'll pre-approve it.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29797