On Thursday, 22 August 2024 16:15:51 BST Georg Gast wrote: > Hi all, > > my target MCU is STM32F103RBT6. A Cortex-M3 with 20KiB RAM and 128Kib > flash. So arm-none-eabi is right. > > This is how i tried to compile it from git src. All make operations are > without -jxx. It took a pretty long time. You'll be glad to hear the script I have runs everything with "NPROC-1" so you'll be to do it faster next time. Ditto with glibc/linux. > The reason why i try to compile it from source is: It was easy to > compile the current gcc from src for x86_64-linux so i wanted to try > this too and learn something on the way. > > Also figured out my fault in my previous attempts: I had to export the > new PATH that newlibs configure can pick up the new arm-none-eabi-gcc. I had that too. I'd forgotten until you mentioned it. For both my newlib and gcc build functions it appears I have done this for the duration of those functions.. p="$PFX" PATH="$p""/bin:""$PATH" LD_LIBRARY_PATH="$p""/lib64:""$p""/lib32:""$p""/lib" export PATH LD_LIBRARY_PATH ..where "PFX" is the install prefix (--prefix=) value. The reason for LD_LIBRARY_PATH is lost to time but looking at that pattern, it's what I use when the wrong library gets found. Maybe I put it in out of habit or maybe some build on some box with some OS barfed on encountering the wrong library.
