Fix FYSETC S6 FLASH_PAGE_SIZE

[sjasonsmith]

Description

The Maple framework used for the STM32F1 HAL defines the type uint32, but the framework used by the STM32 HAL does not.

The S6 variants file was modified to use uint32, which caused the compilation error shown below. I have correct this, and added a test to avoid future regressions, and to provide some coverage of the STM32 HAL.

arduinoststm32\cores\arduino\stm32\stm32_eeprom.c:35:
C:\Users\jason\.platformio\packages\framework-arduinoststm32\variants\FYSETC_S6/variant.h:152:33: warning: implicit declaration of function 'uint32' [-Wimplicit-function-declaration]
 #define FLASH_PAGE_SIZE         uint32(4 * 1024)

Benefits

Fix S6 builds.

Related Issues

I think this was broken only a few days ago, so I don't believe there is an open issue.

[sjasonsmith]

I am not sure why that S6 build is failing in the check. It is building fine for me, both in Windows/VSCode and from the command line in Windows Subsystem for Linux.

There must be a dependency difference or something, but I've not succeeded in narrowing it down yet.

[AnHardt]

Could it be, someone wanted to write uint32_t instead of uint32?

[sjasonsmith]

Could it be, someone wanted to write uint32_t instead of uint32?

That is what I initially thought, but several STM32F1 boards are fine with uint32. That is what led me to track down that maple defines its own types without the ‘_t’.

I didn’t use uint32_t here because nothing else uses stdint types in the variant file, and there is no need for it to be that explicit.

Unfortunately I still have not figured out why my new test fails when run through GitHub. I assume there is something simple I’m doing wrong.

[sjasonsmith]

I was able to reproduce the failure locally if I run the following:

run-tests . FYSETC_S6

After pulling your change to mftest, it now passes locally, so I don't know what is causing the check to continue failing.

[thinkyhead]

The errors are being thrown in framework-arduinoststm32/cores/arduino/ files that are unable to see certain functions or defines. I'm checking to see which version(s) of the platform throw the error and which do not. So far it looks like 6.0.0 is the culprit, so I'll force it to use 5.7.0 (or at least not 6.0.0).

[sjasonsmith]

The errors are being thrown in framework-arduinoststm32/cores/arduino/ files that are unable to see certain functions or defines. I'm checking to see which version(s) of the platform throw the error and which do not. So far it looks like 6.0.0 is the culprit, so I'll force it to use 5.7.0 (or at least not 6.0.0).

I think it has something to do with how the variants files are copied into the .platformio/packages/arduinoststm32/variants/FYSETC_S6 folder.
I found that if I delete the variants that get copied over, the failures return. If I then run the test again, it will pass.

[thinkyhead]

I found that if I delete the variants

I deleted all old platformio files to get a clean slate and platform = ststm32@<6 worked where ststm32@>=6 did not.

…but the Github builder is still picking up something different from the local test.

[sjasonsmith]

I found that if I delete the variants

I deleted all old platformio files to get a clean slate and platform = ststm32@<6 worked where ststm32@>=6 did not.

…but the Github builder is still picking up something different from the local test.

I think I found a problem with the extra script used. Will push a change in a moment.

[sjasonsmith]

I deleted all old platformio files to get a clean slate and platform = ststm32@<6 worked where ststm32@>=6 did not.

I believe I was building fine with 6.0 locally, I suspect it was the issue of the script not using the pre: tag so you may not have had the variant created properly when you tried 6.0.

[thinkyhead]

Yes, on repeat attempts I could no longer get back to the point where the error was. Sounds like you've got the right culprit.