------- Comment #3 from aflyhorse at foxmail dot com 2010-03-23 16:30 ------- (In reply to comment #2) > We need a testcase to reproduce the bug (preprocessed source). You also > might want to try removing -mcmodel=large (you really shouldn't need that).
I've tryed to remove all newly added optimize option seperately, and only remove the -mcmodel doesn't work. The resolving way is to remove all the options concerning sse3 (including -march=k8-sse3 -mtune=k8-sse3 -Wa,march=k8+sse3; use -march=k8 and -nsse3 produce the same error). I am building binutils (SVN 20100317) at that time. I have tried various way of CCFLAGS struggling to avoid this problem. In addition, file: /libcpp/files.c: Line 609-614: comparison is always false under win x64 and /gcc/emit-rtl.c: Line 361: /gcc/ggc-common.c: Line 310: /gcc/graphite-dependences.c: Line 107: /gcc/src/gcc/ira-conflicts.c: Line 125: /gcc/src/gcc/pointer-set.c: Line 67: /gcc/src/gcc/tree dump: Line 168: /gcc/src/gcc/cp/class.c: Line 6740, 6742, 6764, 6767, 6900: all have host-dependent code such as explictly cast a pointer to long (in win x64 means long long to long) and use %ld or &lx to fprintf a pointer. they cause the cc1 treat warnings as error and stop the make procedure, though they are easy to fix manually but quite annoying while making bootstrap. (These are gcc 4.5.0 20100322 version source) -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43477