Closed Baggypants closed 4 years ago
Tag v1.0 compiles and runs.
Ok, I'm just using Fedora packaged arm-none-eabi-gcc-cs-c++-1:9.2.0-1.fc31.x86_64
which does provide math.h in /usr/arm-none-eabi/include/c++/9.2.0/math.h
I tried stuffing -I /usr/arm-none-eabi/include
in to https://github.com/rsta2/circle/blob/d2cc5335229a7ecb889ecd6a8d93a5acde79426e/Rules.mk#L125 ,which I admit was a bit of a punt, but it didn't seem to like it. You can freely chuckle at my ineptitude.
Oh wait, I've just realised there's one in
/arm-none-eabi-gcc-cs-c++-1:9.2.0-1.fc31.x86_64
/usr/arm-none-eabi/include/c++/9.2.0/math.h`
and another in
arm-none-eabi-newlib-3.1.0-4.fc31.noarch
/usr/arm-none-eabi/include/math.h
I'll have to check which one I thought I was including this evening.
Normally the system header files will be found automatically by the compiler, if the option -nostdinc
is not specified, what is the case. I'm using the toolchain from here (gcc-arm-9.2-2019.12-x86_64-arm-none-eabi.tar.xz). It should work in any case with this one.
There is a new branch "test" in the repo. I reverted the commit, which moved the math functions over to the toolchain libraries there and implemented the fabsf() and sqrtf() functions, so that they work with AArch64. I successfully built this version for RPi 1 on Raspbian Buster on a RPi 3B and it runs. It should build on a RPi 4 too.
Oh, well, I just got the toolchain stuff sorted too! It compiles and runs cleanly in master.
No problem! It's good, that it builds now with the recommended toolchain for you. I can easily remove the "test" branch again and it wasn't a big effort.
I have removed the "test" branch. I think, this can be closed.
using tag v2.0 and head.
I'm assuming from having a look in circle that you used circle-stdlib somehow? It's not obvious to me how to include it.