http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47596
Ozkan Sezer <sezeroz at gmail dot com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |sezeroz at gmail dot com --- Comment #6 from Ozkan Sezer <sezeroz at gmail dot com> 2011-02-17 10:12:34 UTC --- The toolchain in question was created by me and I finally found time to look at this. I want to this PR reopened because I _can_ reproduce the ICE with unmodified gcc: gcc-4.4.5 (svn rev. 164871), binutils-2.20.51 (cvs 2010-10-03 08:20 GMT) and mingw-w64 crt and headers (release-1.0 branch, svn rev. 3696), all unmodified, I get this: $ i686-w64-mingw32-g++ -v Using built-in specs. Target: i686-w64-mingw32 Configured with: ../gcc44-svn/configure --target=i686-w64-mingw32 --disable-multilib --enable-checking=release --prefix=/home/ozzie/cross_win32 --with-sysroot=/home/ozzie/cross_win32 --enable-languages=c,c++ --disable-nls Thread model: win32 gcc version 4.4.5 (GCC) $ i686-w64-mingw32-g++ -fstack-check -O gen-duplicator-batch-files.cpp gen-duplicator-batch-files.cpp: In function 'int main(int, char**)': gen-duplicator-batch-files.cpp:610: internal compiler error: in print_reg, at config/i386/i386.c:10678 Please submit a full bug report, with preprocessed source if appropriate. See <http://gcc.gnu.org/bugs.html> for instructions. This is on fedora10 x86_64-linux, the toolchain compiled using the fedora-provided gcc-4.3.2. The problem seems to be related to the -fstack-check switch, without it there is no ICE. Kai: How did you test this?