+ Change how queues are allocated and deleted so only one pvPortMalloc() or vPortFree() is required in place of the previous 2.
+ Where the TCB is allocated in relation to the stack is now dependent on the stack growth direction. The stack will not grow into the TCB.
+ Introduce the configAPPLICATION_ALLOCATED_HEAP constant to allow the application to provide the array used by heap_4.c as its heap. This allows the application writer to use qualifiers on the array to, for example, force the memory into faster RAM.
Demo application:
+ Add demo for SAMA5D4 using IAR.
+ Introduce xSemaphoreGenericGiveFromISR() as an optimisation when giving semaphores and mutexes from an interrupt.
Demo applications:
+ Update IntSemTest.c to provide more code coverage in xSemaphoreGenericGiveFromISR().
+ Ensure the MMU is turned on in the RZ IAR demo. It was already on in the RZ ARM demo.
- Re-introduce the ability to give a mutex from an ISR.
Common demo code:
- Add additional tests into the GenQTest files for priority inheritance and using a mutex from an ISR.
Add additional asserts to the Keil CM3 and CM4F ports (other CM3/4 ports already updated).
Add the additional yield necessitated by the mutex held count to the case when configUSE_QUEUE_SETS is 0.
Helper updates to allow a count of the number of mutexes held to be added.
Updates to the CCS Cortex-R4 implementation necessitated by a change in compiler semantics.
Update PIC32MX and MZ ports to assert if a non ISR safe function is called from an ISR.
Fix issue in and simplify the xTaskRemoveFromUnorderedEventList() function. The function is new to the V8 release candidates so does not effect official released code.
Replace LDMFD with POP instructions in IAR and RVDS Cortex-A9 ports.
Replace branch to address with indirect branch and exchange to address in register in the IAR and RVDS Cortex-A9 ports.
Change INCLUDE_xTimerPendFunctionCallFromISR to INCLUDE_xTimerPendFunctionCall
Update event group trace macros to match the new trace recorder code.
Ensure parameter name consistency by renaming any occurrences of xBlockTime and xBlockTimeTicks to xTicksToWait.
Continue work on GCC/RL78 port - still a work in progress.
Adjust how the critical section was used in xQueueAddToSet.
Remove obsolete extern declaration of vTaskSwitchContext() from the MPX430X IAR portmacro.h (other older portmacro.h header files contain the same declaration).
When a task is unblocked the need for a context switch is only signalled if the unblocked task has a priority higher than the currently running task, instead of higher than or equal to the priority of the currently running task.
Replace #define types with typedefs.
Rename all typedefs to have a _t extension.
Add #defines to automatically convert old FreeRTOS specific types to their new names (with the _t).
Introduce xPortRunning variable into Win32 simulator port layer.
Add port optimised task selection macro for the GCC Win32 port layer (the MSVC version has had one for a while).
Ensure the event list item value does not get modified by code in tasks.c (priority inheritance, or priority change) when it is in use by the event group implementation.
Reset the expected block time variable when a task is suspended or deleted in case the value held in the variables was associated with the task just suspended or deleted.
Add a task pre-delete hook to allow the insertion of any port specific clean up when a task is deleted.
Increase use of 'const' qualifiers.
Add vPortCloseRunningThread() into the Win32 port layer to attempt to allow Windows threads to be closed more gracefully when a task deletes itself.
Correct header comments in the UARTCommandConsole.c file used in the SmartFusion2 demo.
Exercise the new xQueueSpacesAvailable() function in the MSVC demo.
Add defaults for the new traceMALLOC and traceFREE trace macros.
Catch tasks trying to exit their functions in the Cortex-M0 ports.
Add additional comments to timers.c in response to a support forum question.
Initialise _impure_ptr prior to the first task being started.
Prior to V7.5.0 a yield pended in the tick hook would have occurred during the same tick interrupt. Return pdTRUE from xTaskIncrementTick() if a yield is pending to revert to that behaviour.
Relegate the vTaskList() and vTaskGetRunTimeStats() functions to "sample" functions demonstrating how to use xTaskGetSystemState() to generate human readable status information.
Introduce and default configINCLUDE_STATS_FORMATTING_FUNCTIONS which must now be defined to use vTaskList() and vTaskGetRunTimeStats().
Add eTaskConfirmSleepModeStatus() (not yet tested).
Only call prvQueueUnregisterQueue() when a queue or semaphore is deleted if configQUEUE_REGISTRY_SIZE > 0.
Back out change that checks the configUSE_PORT_OPTMISED_TASK_SELECTION value before uxPriorityUsedOnEntry is set in vTaskPrioritySet as it generated more warnings (with other compilers) than it fixed.
Made queue function prototypes consistent so xQueueHandle parameters are always xQueue, and xQUEUE * parameters pxQueue.
Likewise make the task API using px for pointers to TCBs, and just x for task handles.
Heap_x functions now automatically align the start of the heap without using the portDOUBLE union member.
Queue.c now includes queue.h.
Changed eTaskStateGet() to eTaskGetState() and added #define to ensure backward compatibility.
Added configEXPECTED_IDLE_TIME_BEFORE_SLEEP definition - was previously hard coded to 2.
Slight change to the default CM3 tickless sleep function to allow the idle time to be set to zero in the pre-sleep processing macro.
Changed stack alignment for the FreeRTOS-MPU port to ensure it didn't trigger the assert() in the generic create function.
Move location of traceTASK_CREATE() macro call.
Remove obsolete handling of trmCOMMAND_PROCESS_TIMER_OVERFLOW as the command was never used in release versions.
Replaced the #error that traps configMAX_SYSCALL_INTERRUPT_PRIORITY being set to 0 with a configASSERT() for GCC Cortex-M3/4 ports as the #error does not work if configMAX_SYSCALL_INTERRUPT_PRIORITY includes any casting. Not a problem for other compilers as they cannot have casting anyway as that would break the assembly code.
Richard Barry