boards/xg23-pk6068a: remove HFXO_FREQ and LFXO_FREQ

[basilfx]

Contribution description

Prior to the use of Gecko SDK 4.5, RIOT shipped with patched vendor files, such as system_xxx.c. These system files included periph_config.h, which allowed one to redefine HFXO_FREQ and LFXO_FREQ.

This was not intended, and other EFM32-based boards documented this to redefine this using compiler defines. After all, a board configuration is considered to be somewhat pre-defined and static.

With the new Gecko SDK 4.5, this (unintended) feature got lost for the xg23-pk6068a board. To bring it in-line with the other boards, remove the defines, and document how it should be used. Do note that, for some reason, Silicon Labs decided not to prefix the aforementioned defines with EFM32_.

Within the codebase, LFXO_FREQ was used for ztimer configuration. Similar to the other boards, this has been replaced with a static value.

Behind the scenes, the vendor system_xxx.c file will still use the same values.

Testing procedure

The compiler defines are equal to the values used in system_efr32zg.c, so there is effectively no change if you don't override these values. See https://github.com/basilfx/RIOT-gecko-sdk/blob/master/dist/platform/Device/SiliconLabs/EFR32ZG23/Source/system_efr32zg23.c.

I do not own this board, so I cannot test if the documented alternative works.

Issues/PRs references

Regression introduced in #22040
Fixes #22283

Declaration of AI-Tools / LLMs usage:

AI-Tools / LLMs that were used are:

• none

[riot-ci]

Murdock results

✔️ PASSED

dc9133c dist/tools/doccheck: remove warnings

Success	Failures	Total	Runtime
554	0	554	05m:37s

Artifacts

• Documentation preview

[basilfx]

The only validation I can do, is to see if the variables end up at the right place. That's the case:

$ CFLAGS="-DHFXO_FREQ=notanint -DLFXO_FREQ=123456789123456789" BOARD=xg23-pk6068a make -j14
Building application "default" for "xg23-pk6068a" with CPU "efm32".

"make" -C pkg/cmsis/
"make" -C pkg/gecko_sdk/
"make" -C build/pkg/gecko_sdk/dist
"make" -C build/pkg/gecko_sdk/dist/platform/common
"make" -C build/pkg/gecko_sdk/dist/platform/emlib
"make" -C build/pkg/gecko_sdk/dist/platform/emlib-extra
"make" -C pkg/mpaland-printf/
"make" -C build/pkg/mpaland-printf -f pkg/mpaland-printf/mpaland-printf.mk
"make" -C boards
"make" -C boards/xg23-pk6068a
"make" -C core
"make" -C core/lib
"make" -C cpu/efm32
"make" -C drivers
"make" -C sys
"make" -C boards/common/init
"make" -C sys/auto_init
"make" -C drivers/periph_common
"make" -C sys/div
"make" -C drivers/saul
"make" -C sys/fmt
"make" -C drivers/saul/init_devs
"make" -C sys/isrpipe
"make" -C sys/libc
"make" -C sys/malloc_thread_safe
"make" -C sys/newlib_syscalls_default
"make" -C sys/phydat
"make" -C sys/pm_layered
"make" -C sys/preprocessor
"make" -C sys/ps
"make" -C sys/saul_reg
"make" -C cpu/cortexm_common
"make" -C sys/shell
"make" -C sys/shell/cmds
"make" -C sys/stdio
"make" -C sys/stdio_uart
"make" -C sys/tsrb
"make" -C cpu/cortexm_common/periph
"make" -C cpu/efm32/families/efr32zg23
"make" -C cpu/efm32/periph
In file included from <command-line>:
examples/basic/default/bin/xg23-pk6068a/riotbuild/riotbuild.h:18:19: error: unsigned conversion from 'long long int' to 'uint32_t' {aka 'long unsigned int'} changes value from '123456789123456789' to '2899336981' [-Werror=overflow]
   18 | #define LFXO_FREQ 123456789123456789
      |                   ^~~~~~~~~~~~~~~~~~
build/pkg/gecko_sdk/dist/platform/Device/SiliconLabs/EFR32ZG23/Source/system_efr32zg23.c:98:35: note: in expansion of macro 'LFXO_FREQ'
   98 | static uint32_t SystemLFXOClock = LFXO_FREQ;
      |                                   ^~~~~~~~~
build/pkg/gecko_sdk/dist/platform/Device/SiliconLabs/EFR32ZG23/Source/system_efr32zg23.c: In function 'SystemMaxCoreClockGet':
examples/basic/default/bin/xg23-pk6068a/riotbuild/riotbuild.h:17:19: error: 'notanint' undeclared (first use in this function)
   17 | #define HFXO_FREQ notanint
      |                   ^~~~~~~~
build/pkg/gecko_sdk/dist/platform/Device/SiliconLabs/EFR32ZG23/Source/system_efr32zg23.c:438:31: note: in expansion of macro 'HFXO_FREQ'
  438 |   return(HFRCODPLL_MAX_FREQ > HFXO_FREQ \
      |                               ^~~~~~~~~
examples/basic/default/bin/xg23-pk6068a/riotbuild/riotbuild.h:17:19: note: each undeclared identifier is reported only once for each function it appears in
   17 | #define HFXO_FREQ notanint
      |                   ^~~~~~~~
build/pkg/gecko_sdk/dist/platform/Device/SiliconLabs/EFR32ZG23/Source/system_efr32zg23.c:438:31: note: in expansion of macro 'HFXO_FREQ'
  438 |   return(HFRCODPLL_MAX_FREQ > HFXO_FREQ \
      |                               ^~~~~~~~~
cc1: all warnings being treated as errors
make[3]: *** [Makefile.base:162: examples/basic/default/bin/xg23-pk6068a/cpu_efr32zg23/build/pkg/gecko_sdk/dist/platform/Device/SiliconLabs/EFR32ZG23/Source/system_efr32zg23.o] Error 1
make[3]: *** Waiting for unfinished jobs....
make[2]: *** [Makefile.base:31: ALL--cpu/efm32/families/efr32zg23] Error 2
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [Makefile.base:31: ALL--cpu/efm32] Error 2
make: *** [examples/basic/default/../../../Makefile.include:743: application_default.module] Error 2

[basilfx]

Rebased to see if this fixes the build sanity checks:

  	Update manifest digest to 84746b38b9a6248a21bfa2f9bd4d7a052870c35a689d9b32a06be2f5d7042156:
  		makefiles/docker.inc.mk:8:DOCKER_TESTED_IMAGE_REPO_DIGEST := b932e0079de853104bdb579a35d5a02d8613c87dc30130c67b03aa3b2027fd26

[crasbe]

No, merging this would fix it though: #22307

I merged some PRs for the RIOT docker container.

[crasbe]

And since #22255 was merged, you don't even have to rebase afterwards. Just rerun the static test job.