> Maybe in the distant past of GCC there were some problems > with this "combined build of GCC and newlib" but someone > suggesting a separate build somewhere nowadays is really weird. It's not that weird. Many people are using that method: 1) The OpenWrt build system does. 2) The ArchWiki describes that way too: https://wiki.archlinux.org/index.php/Cross-compiling_tools_package_guidelines Look for "stage 1" in that page. 3) crosstool-NG is specifically designed to build cross-compilers, and uses the same method: https://crosstool-ng.github.io/docs/toolchain-construction/ Look for "core pass 1" in that page. > [...] > https://docs.rtems.org/releases/4.5.1-pre3/rtemsdoc/html/started/started00072.html I would not trust the instructions on that page anymore. It's way too old. > "To bring in newlib and libgloss, a user could link the newlib and > libgloss subdirectories of the newlib source tree into the src tree of gcc." > In : > https://sourceware.org/newlib/faq.html I would not trust the Newlib documentation very much. Building Newlib for a Linux system (as opposed to the usual embedded system) was broken for years, and it may still be. I would say the project is short on developers. I did find some information in GCC's own Wiki: https://gcc.gnu.org/wiki/Building_Cross_Toolchains_with_gcc But then it says "Create a tree that consists of all of the files from the GCC and binutils/gdb/newlib source trees (including several simulators in src/sim), with the GCC files overriding the binutils/gdb/newlib files when there's a conflict." Overwriting files like that doesn't exactly inspire confidence. I have invested many hours of work to get a working, tailor-made cross-compiler for my embedded projects. I would even say "wasted a lot of time". I hope you understand that, before investing yet more work, I would like to see some modern, working example (makefile or script) first on how to build a GCC cross-compiler with Newlib in a single step. Regards, rdiez
