FreeRTOS v8.1.0 on 64-Bit Platforms

pmendham wrote on Wednesday, August 27, 2014:

We are building FreeRTOS v8.1.0 for 64-bit Linux using the POSIX simulator. This is almost all fine, the main addition being the inclusion of a definition for portPOINTER_SIZE_TYPE in portmacros.h:

#define portPOINTER_SIZE_TYPE uintptr_t

(If you see something wrong in this, please say!) However, a single warning remains which, to me, looks like a legitimate concern. On line 556 of tasks.c there is an assert:

configASSERT( ( ( ( uint32_t ) pxTopOfStack & ( uint32_t ) portBYTE_ALIGNMENT_MASK ) == 0UL ) );

In order to silence the warning, we have changed this to:

configASSERT( ( ( ( portPOINTER_SIZE_TYPE ) pxTopOfStack & ( portPOINTER_SIZE_TYPE ) portBYTE_ALIGNMENT_MASK ) == 0U ) );

Given that the line above this (555) has a documented MISRA exception on it, and carefully makes use of portPOINTER_SIZE_TYPE, why does the assert cast to uint32_t? Is this a problem for anyone else (e.g. on the Windows simulator), or are we doing something differently/wrong?

Whatever the solution on this line, the same needs to be applied to line 563, which is the same assert, but for positive stack growth.



rtel wrote on Wednesday, August 27, 2014:

Probably just because the asserts have mostly been used on devices with 32-bit pointers, or where the pointer is less than 32-bits so casting to a uint_32 is fine. There is no deliberate reason why. We can change it if it causes you an issue.