Unconditionally define POSIX sysconf macros PAGE_SIZE and PAGESIZE

[besmarsh]

The POSIX sysconf() option _SC_PAGESIZE was defined as PAGESIZE, which was defined as PAGE_SIZE (coming from CONFIG_POSIX_PAGE_SIZE) IF not already defined. The STM32 HAL defines PAGESIZE for flash pages as FLASH_PAGE_SIZE, which is only defined in the HAL for some STM32 families. This caused build failures for many STM32 targets. See issue #74623 for full description.

In addition, the POSIX sysconf() options _SC_PAGE_SIZE and _SC_PAGESIZE should be synonymous, but the existing definition could result in them differing when PAGESIZE is already defined elsewhere.

This issue was present in v3.6 but is more of an issue for v3.7 as it causes build failures for applications not even using sysconf at all, if they target an affected STM32 family and enable CONFIG_POSIX_API and CONFIG_CPP.

This PR removes the conditional definition of PAGESIZE and uses PAGE_SIZE (set from CONFIG_POSIX_PAGE_SIZE) for both _SC_PAGE_SIZE and _SC_PAGESIZE.

The POSIX macros PAGE_SIZE and PAGESIZE (queriable through sysconf()) were conditionally defined only if an existing definition did not already exist. These should be defined unconditionally in their header to ensure they get the correct values.

If these macros are defined elsewhere with a different meaning, that's a problem. There was an issue where PAGESIZE was already defined with a different meaning. See #74623 and #74428.

The POSIX macro ATEXIT_MAX is also conditionally defined and should be unconditionally defined, but there is currently a definition in picolibc (picolibc/newlib/libc/include/stdlib.h) so this change will be done separately.

This PR defines PAGE_SIZE and PAGESIZE unconditionally.

[besmarsh]

I've just seen #70136, #73363, and #74428 regarding this same issue. The approach there was to make changes to the STM32 HAL - I guess that solves the issue too, but I think the fix here is good for making the _SC_PAGE_SIZE and _SC_PAGESIZE options consistent and equal.

[cfriedt]

#74428 should fix this

[besmarsh]

This is fixed by #74428, but this change ensures that no other rogue PAGESIZE definitions can cause the same issue, and that _SC_PAGE_SIZE and _SC_PAGESIZE always return the same.

[cfriedt]

Removing the conditions around defining PAGESIZE should be all that is needed here, but PAGESIZE also needs to be defined according to the spec.

https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/limits.h.html

[cfriedt]

Suggested change

	#ifndef PAGESIZE
	#define PAGESIZE PAGE_SIZE
	#endif
	#define PAGESIZE PAGE_SIZE

[besmarsh]

Should the conditions around the definitions of PAGE_SIZE and ATEXIT_MAX just above also be removed to ensure they are defined correctly?

[cfriedt]

Yes, please

[cfriedt]

It would be preferable to keep the original definition here. Otherwise, it would be an unnecessary special case for scripts.

[besmarsh]

Keeping the original definition is fine if PAGESIZE and PAGE_SIZE are unconditionally defined equal.

[cfriedt]

Yes

[besmarsh]

PR updated, including change of title and description after scope change.

[besmarsh]

ATEXIT_MAX is also defined in picolibc/newlib/libc/include/stdlib.h - I don't know if it should be there or not

[cfriedt]

@besmarsh - if you prefer, you can go back to the way it was before.

There will be a "Great Reckoning" not far down the road where we make internal Zephyr headers as consistent as possible with both Picolibc and Newlib.

[erwango]

Removing the STM32 label as this part of the issues has been solved

[besmarsh]

@besmarsh - if you prefer, you can go back to the way it was before.

There will be a "Great Reckoning" not far down the road where we make internal Zephyr headers as consistent as possible with both Picolibc and Newlib.

I've removed the ATEXIT_MAX change and just changed PAGE_SIZE and PAGESIZE. The ATEXIT_MAX change can wait for the alignment of Zephyr and Picolibc/Newlib headers.