Dave Yost wrote: > I'm trying to build a 4.1.1 on cygwin that will let me use the > -mno-cygwin argument. I looked and looked on the web for the runes > and finally heard somewhere in a dark alley that I should try > something like the below. I first ran Cygwin setup, and everything > is up to date. I think you're a little confused here. There are two main gcc targets on win32, Cygwin and MinGW. The Cygwin gcc target has a -mno-cygwin option which is just a shortcut for invoking the MinGW gcc if it is installed in the right place (i.e. in a parallel directory structure.) If you want to compile things with the MinGW gcc then you'll obviously have to build that, but if you want to invoke the MinGW gcc as "gcc -mno-cygwin" then you'll need to also build the native Cygwin gcc of the same version and install them both in the same place. It seems like a lot of work to go through to build two compilers when you could just build a native MinGW gcc and invoke it directly. > Z% ./configure \ > --prefix=/usr/local \ > --enable-version-specific-runtime-libs \ > --disable-nls \ > --host=i686-pc-cygwin \ > --target=i686-pc-mingw32 Here you are cross compiling (in the eyes of autoconf at least) even though both targets are nearly compatible at the assembler and linker level. > echo timestamp > stmp-int-hdrsmake[2]: *** No rule to make target > `/usr/local/bin/i686-pc-mingw32-as.exe', needed by `stamp-as'. Stop. But here autoconf can't find the assembler for the target you specified. It doesn't know that it can use the native host assembler, because it thinks you are cross compiling. Which you are, technically. So you'll need to build a cross-binutils, or at least provide symlinks or shell wrappers so that they are available under the target-prefixed alias. Brian
