boards/arm/gcc: disable coverage analysis temporarily

[anchao]

Summary

boards/arm/gcc: disable coverage analysis temporarily

Please revert this change if arm resolved this issue in the next release

ARM GNU toolchain 11.3.Rel1 missing gcov_* symbols:

https://community.arm.com/support-forums/f/compilers-and-libraries-forum/53354/arm-gnu-toolchain-11-3-rel1-missing-gcov_-symbols

$ arm-none-eabi-gcc -v

...
gcc version 11.3.1 20220712 (Arm GNU Toolchain 11.3.Rel1)

$ cat main.c

int __start(void)
{
  return 0;
}

Compiler with arm GCC-11.3:

$ arm-none-eabi-gcc -c -Wstrict-prototypes -Os -fno-strict-aliasing -fomit-frame-pointer -fprofile-generate \ -ftest-coverage -mthumb --coverage -fno-common -Wall -Wshadow -Wundef -ffunction-sections -fdata-sections -g \ -mlittle-endian -mcpu=cortex-m0 -mthumb -mfloat-abi=soft main.c -o main.o

$ arm-none-eabi-ld --entry=__start -nostdlib --gc-sections -Tpimoroni-tiny2040-flash.ld -o test_gcov --start-group main.o \ ~/gcc/linux/arm/bin/../lib/gcc/arm-none-eabi/11.3.1/thumb/v6-m/nofp/libgcov.a --end-group arm-none-eabi-ld: main.o: in function `__start':
/home/archer/code/nuttx/n4/incubator-nuttx/main.c:2: undefined reference to `__gcov_indirect_call_profiler_v4' 
arm-none-eabi-ld: /home/archer/code/nuttx/n4/incubator-nuttx/main.c:4: undefined reference to `__gcov_indirect_call' 
arm-none-eabi-ld: /home/archer/code/nuttx/n4/incubator-nuttx/main.c:4: undefined reference to `__gcov_time_profiler_counter'

link success with arm GCC-10.3:

$ arm-none-eabi-ld --entry=__start -nostdlib --gc-sections -Tpimoroni-tiny2040-flash.ld -o test_gcov --start-group main.o \ ~/lib/gcc/arm-none-eabi/10.3.1/thumb/v6-m/nofp/libgcov.a --end-group
$ ls -l test_gcov
-rwxrwxr-x 1 archer archer 68744 10月 24 12:51 test_gcov

Signed-off-by: chao an anchao@xiaomi.com

Impact

$ arm-none-eabi-gcc -v
...
gcc version 11.3.1 20220712 (Arm GNU Toolchain 11.3.Rel1)

Testing

CI- check

[pkarashchenko]

Anyway I still think that we are too fast with toolchain upgrade, but I'm more conservative. Do we have a significant gain in optimisation with new compiler? Or what is the rush here?

[anchao]

Anyway I still think that we are too fast with toolchain upgrade, but I'm more conservative. Do we have a significant gain in optimisation with new compiler? Or what is the rush here?

The new toolchain adds more static analyzer on compile stage, which will helpful for code quality:
https://gcc.gnu.org/gcc-11/changes.html

For example:

    if (focus != NULL) {
        nxipc_client_disconnect(focus->send_proxy);
        nxipc_sub_disconnect(focus->receive_proxy);
        free((void*)focus->name);
        free((void*)focus);
        pthread_mutex_destroy(&focus->mutex);   <--  use-after-free
    }

build will break on gcc-11.3 ([-Werror=use-after-free])

../common/src/client/app_focus_c.c: In function 'app_focus_release':
../common/src/client/app_focus_c.c:157:9: error: pointer 'handle' used after 'free' [-Werror=use-after-free]
  157 |         pthread_mutex_destroy(&focus->mutex);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../common/src/client/app_focus_c.c:156:9: note: call to 'free' here
  156 |         free((void*)focus);
      |         ^~~~~~~~~~~~~~~~~~

[xiaoxiang781216]

Let's merge this patch to unblock ci.