Okay, found the issue with the crash (I think). It looks like: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=8049 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=10395
The relevant instructions, when single-stepping through sprintf(), are: 0x00002b008c63555f <sprintf+95>: mov %rsp,%rdx 0x00002b008c635562 <sprintf+98>: jmpq *%r8d <<movaps instructions that aren't hit>> 0x00002b008c635581 <sprintf+129>: movaps %xmm0,0xffffffffffffff81(%rcx) As any/all floating point values are handled in *printf() with xmm registers, the first movaps instruction causes it to spit out a segfault. I'm a bit fuzzy on the full implications, but if I'm reading it right (and eyeballing the disassembly right), it needs glibc rebuilt, too. Looks like gprolog on amd64 can't be built as a native binary until glibc is rebuilt and a backport for gcc 3.4.x is done, or gprolog gets ported to gcc 4.x. Argh. :( Or I could be completely and totally wrong. -- DN Daniel -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
