MP1 support

[Inujel]

This is still very much a work in progress.

I create the PR because I'd like to discuss a few points before going further 🙂.

Could you please give me some feedback about the comments below?
Thank you very much!

[Inujel]

It looks like for the MP1, CMSIS is dependant on HAL : system_stm32mp1xx.c includes stm32mp1xx_hal.h

CMSIS cannot be used without HAL for the MP1. I don't think that stm32-cmake should do anything special about it.
The CMSIS test cannot succeed. I remove MP1 from the default list and add a warning about it.

if(NOT TEST_FAMILIES)
    set(TEST_FAMILIES F0 F1 F2 F3 F4 F7 G0 G4 H7 L0 L1 L4 L5 U5 WB WL)
endif()

if("MP1" IN_LIST TEST_FAMILIES)
    message(WARNING "CMSIS for MP1 devices requires HAL, this test is expected to fail for the MP1 family")
endif()

[Inujel]

Hi @atsju

It's ready to be reviewed.

[atsju]

you did not manage it like described in our previous discussion. I think you did not get how MP1.cmake is supposed to work (I know it's complex as I added several families in the past). F3.cmake might be my best guess to look at. And look at how stm32_get_chip_type is working

Let me know if you want to dig or if you want me to have a look. It's not easy for me either.

[Inujel]

Sorry about that. I remember changing it. I guess I screwed up something when committing 😅 .
This is fixed.

I admit that I did not understand everything about types and families.
I will look into stm32_get_chip_type, but can you tell me what feels wrong to you?

[atsju]

as discussed previously the correct startup could be used instead the generic one.

[atsju]

I think this is not completely correct. all size codes are incorrect for MP1 family. It must be managed explicitly and not just removing a warning.

[Inujel]

There's no flash on the MP1 family. The only size code is "A" and I though that's what it meant. Don't you agree?
I did not put that there just to remove a warning.

To me, this part just don't apply for the MP1 family.

[atsju]

in STM32G473RE the E is the size code.
Please also have a look at datasheet. From my point of view, MP1 has no such thing as a size_code.
maybe you should do same as L1 family (see stm32l1_get_memory_info) to return always 0.
All these information makes me think a modification is needed.

[atsju]

We are almost done. Seems very promising.

[Inujel]

I added BSP support for the discovery and evaluation boards. However the bsp test currently fails.
I think it's the case for all multicore MCUs because the test target depends on BSP::STM32::${BOARD} instead of BSP::STM32::${BOARD}${CORE_C}.

CMake Error at CMakeLists.txt:59 (add_executable):
  Target "bsp-test-STM32WB15CC_Nucleo" links to target
  "BSP::STM32::STM32WB15CC_Nucleo" but the target was not found.  Perhaps a
  find_package() call is missing for an IMPORTED target, or an ALIAS target
  is missing?

I propose we fix this in a separate PR.

[atsju]

I propose we fix this in a separate PR.

Please just clean the "size code" thing and fix the startup script.
For the BSP test we can indeed wait a bit more.

[Inujel]

There's still something wrong in the fetch test but I can't reproduce it on my machine. I'm searching...

[atsju]

For me this is good to merge. If you feel like BSP will be fixed later then remove it from the tests so that it doesn't fail .
I let you fix the fetch also.

@Hish15 we will need your review for this

After merge will open an issue for BSP test not working and CMSIS test that needs ST correction.

[Hish15]

Good job!

[atsju]

Yes well done! This was not an easy task and you were not the first trying out.
Also you asked good questions and had excellent proposals. Thank you.

[Inujel]

Thanks for your kind words!