> That seems like an unnecessary worry. Why would the > phase 1 compiler install *more* than the final one? That is a very bold claim to make. GCC is a complex beast. Are you sure about that? Is there any test case for it? How about in the future? What if only need the C compiler to build Newlib, but the final toolchain only needs the C++ compiler, in order to reduce the total toolchain build time and save space in my final container? Installing the phase 1 GCC into a separate, temporary bin directory is a clean solution, and I think it is a reasonable goal to pursue. In any case, I think the answers to the following questions are still interesting to know, as they may apply to other scenarios too: 1) Is there a way to convince GCC to find the Binutils with their full names (like arm-none-eabi-ar) in the PATH or with --with-build-time-tools=/some/dir , instead of just short names like "ar"? 2) Are --with-as and --with-ld always risky in this scenario, because of other tools like objdump, given that GCC's own documentation warns about an inconsistent set of Binutils? Regards, rdiez
