Compile C++ code on CWLITE XMEGA

1

Hello,

I am trying to compile a c++ code CWLITE XMEGA, and whenever I start I get this error. The platform & Hal type are not defined. Can you please advise?
…/./hal/hal.h:94:7: warning: “HAL_TYPE” is not defined, evaluates to 0 [-Wundef]
94 | #elif HAL_TYPE == HAL_xmega
| ^~~~~~~~
In file included from FastInf.cpp:4:
…/./hal/hal.h:163:5: warning: “PLATFORM” is not defined, evaluates to 0 [-Wundef]
163 | #if PLATFORM == CW308_MEGARF

2

If you’re using our build and HAL infrastructure, then you need to define PLATFORM (look at any make command in our Jupyter tutorials for reference).
The supported platforms are defined in hal/Makefile.hal.

3

I did define the PLATFORM in cell 1 and in the make command just like all the other tutorials, but it still gave the same errors.

4

image
adding these lines in the make files made the compilation work, but is there any other way to make it work without manually adding these definitions? Also, whenever I add the trigger setup to get the trigger count, I get these linking errors. Can you please advise?
Thanks

image
image770×219 9.65 KB

5

It’s hard for me to debug changes you’ve made on your own modified codebase. Somehow you must be missing our defines in your modified build process. Trace back where/why that happens, using our unmodified build process as a reference. Your linking error is likely tied to this as well.

6

I copied a working c tutorial into a new folder and changed the file name to cpp, so v no modifications to the build process occured. However, I still got the same error.

7

AFAIK, we don’t use any C++ code in any of our projects, so that part of the build system has received little testing. I’d recommend wrapping any header files you include in extern C {} and compiling with make ... VERBOSE=TRUE to see the full build command.

Alex

8

This is what I get using verbose=True

image
image772×465 60.3 KB

image
image707×493 74.1 KB

image
image707×337 49.3 KB

9

Adding on that when I tried to add manually platform definitions in make file, I got this linking error. When I used extern C, the problem still presists.

image
image825×373 56.5 KB

10

Alex can you please check the following and kindly advise because I am using this board for my graduation project

11

Looking into the makefiles more, adding support for C++ will be kind of tricky. As I said earlier, we don’t use C++ in any of our projects, so I’m not sure when this issue popped up, or if C++ ever worked.

I’ll try adding C++ support, but I’m not sure when this will be finished. In the meantime, I recommend trying to convert C++ files you have into C.

Alex

12

Alright, I’ve got something that seems to mostly work on the latest commit. To build C++ files, you’ll need to:

  1. Go to hal/your_platform and modify the makefile inside to define the appropriate CXX compiler. For example, for the F3, I went to hal/stm32f3/Makefile.stm32f3 and added CXX=arm-none-eabi-g++. Usually this will be the C compiler with g++ replacing gcc
  2. In your C++ file, any C headers you #include will need to be wrapped in extern "C" {}. See simpleserial-aes/simpleserial-aes.cpp for an example of this.
  3. In the makefile for your project, add any C++ source files as CPPSRC += my_cpp_source.cpp. See simpleserial-aes/makefile for a commented out example of this

Some notes:

  • CPPFLAGS is just CFLAGS, so there’s no way to do different CPP and C flags
  • The linker used is based on whether or not CPPSRC is empty, so it should use $(CXX) for linking when there’s at least one file in CPPSRC and $(CC) otherwise
13

Thank you so so much Alex for your support, it worked on the STM32F3.
However when I try to use it on the atxmega I get this error. Can you please support

image
image742×175 21 KB

14

You need to use avrg++, not g++

15

Thanks a lot, it worked with CXX= avr-g++