cmake:fix protected build link option break,init canmv230 protected build

[xuxin930]

Summary

this should resolve #12019 (comment)

the link_option of nuttx_user target should be allowed to be set freely

Impact

protected mode

Testing

cmake -B build -DBOARD_CONFIG=canmv230/pnsh

cmake --build build 

[acassis]

@xuxin930 @xiaoxiang781216 any idea why the CI reported this issue many times:

Check failure on line 51 in include/nuttx/wdog.h
GitHub Actions / macOS (macos)

missing binary operator before token "("

[simbit18]

@acassis @xiaoxiang781216 The problem is here
https://github.com/apache/nuttx/actions/runs/8479414330/job/23233475816?pr=12022#step:7:1313
Configuring...
make: *** [olddefconfig] Error 1
Enabling CONFIG_ARM_TOOLCHAIN_GNU_EABI
make: *** [olddefconfig] Error 1

I think clearing the cache solves the problem
https://github.com/apache/nuttx/actions/caches

[xuxin930]

@acassis @xiaoxiang781216 The problem is here https://github.com/apache/nuttx/actions/runs/8479414330/job/23233475816?pr=12022#step:7:1313 Configuring... make: *** [olddefconfig] Error 1 Enabling CONFIG_ARM_TOOLCHAIN_GNU_EABI make: *** [olddefconfig] Error 1

I think clearing the cache solves the problem https://github.com/apache/nuttx/actions/caches

@acassis @simbit18
It seems that similar situations have happened before
refer to: #10543

[simbit18]

@xuxin930 exactly !!! cache must be cleared manually.

[xiaoxiang781216]

why not skip adding these option instead

[xuxin930]

why not skip adding these option instead

These link options is set in the Toolchain file and has global scope.
they are also required by the nuttx target.but the timing of setting the Toolchain file is before all targets.
so it cannot be distinguished by USER_LINK_OPTION and NUTTX_LINK_OPTION like Makefile.

[xiaoxiang781216]

why these options can't be used with uerspace.bin

[yf13]

I don't know how @xuxin930 identified them. My test shows that the gc-sections will cause user space ELF lacking things required by kernel side, thus can't boot the target. The -Map=nuttx.map name conflicts with that of kernel. As these global settings are for kernel, adjusting them for userspace looks reasonable.

[xiaoxiang781216]

@xuxin930 let's split the hack to a new pr

[yf13]

@xuxin930

I tried this and it works, the artifact runs on CanMV230 device as well:

ABC                                                                             
NuttShell (NSH) NuttX-12.4.0                                                    
nsh> cat /proc/version                                                          
NuttX version 12.4.0 da63218c4d Apr  5 2024 08:03:09 canmv230/pnsh
nsh> free                                                                       
                   total       used       free    maxused    maxfree  nused  nfree                                                                              
        Kmem:    1034232       5816    1028416       6408    1028368     16     2                                                                              
        Umem:    1045112       6040    1039072       6040    1039072      9     1 

Only one minor issue met during CMake configuration process:

$ rm -rf * && cmake ~/Projects/Nuttx/nuttx -DBOARD_CONFIG=canmv230/pnsh -G Ninja 
-- Initializing NuttX
--   Board:  canmv230
--   Config: pnsh
--   Appdir: /home/yf/Projects/Nuttx/nuttx/../apps
-- The C compiler identification is GNU 10.2.0
-- The CXX compiler identification is GNU 10.2.0
-- The ASM compiler identification is GNU
-- Found assembler: /usr/bin/riscv64-unknown-elf-gcc
riscv64-unknown-elf-gcc: error:  -Wstrict-prototypes -fno-common -Wall -Wshadow -Wundef -Wno-attributes -Wno-unknown-pragmas: No such file or directory
-- Configuring done
-- Generating done
-- Build files have been written to: /home/yf/tmp/bbb

[yf13]

@xuxin930,

Here I tried this again with GCC toolchain v10.2 got via apt install gcc-riscv64-unknown-elf on Ubuntu.

It seems that we can avoid the above cmake config time warning by dropping the CMAKE_C_FLAGS from the libgcc checking command in line 755.

[xiaoxiang781216]

@xuxin930 should we rebase the code

[xuxin930]

@xuxin930 should we rebase the code

Done, rebased

[xuxin930]

CI seems failed because fetch timeout
try retrigger