------- Comment #2 from cnovikov at pacbell dot net 2007-01-14 06:17 ------- (In reply to comment #1) > Confirmed, not a regression. >
You set keywords to "ice-on-invalid-code". In my opinion, the code is valid. %z is supposed to add an appropriate data width IA-32 insn suffix: l for dwords, w for words, b for bytes. The expected behavior here is for %z0 to produce "b" in the assembler output, since "a" is a char. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30413