GitHub/qpcpp
1
0
Fork 0
mirror of https://github.com/QuantumLeaps/qpcpp.git synced 2025-01-14 05:42:57 +08:00
Code Issues Projects Releases Wiki Activity
qpcpp/doxygen/history.dox
QL b90f1c648c 6.3.1
2018-05-22 15:32:27 -04:00

2288 lines
130 KiB
Plaintext
Raw Blame History

namespace QP {
/** @page history Revision History
@section qpcpp_6_3_1 Version 6.3.1, 2018-05-24
This release migrates the [QUTest](https://www.state-machine.com/qtools/qutest.html) examples to [QM 4.2.1](https://www.state-machine.com/qm/history.html#qm_4_2_1), which now can generate QS_FUN_DICTIONARY() records automatically. This release also adds a generic, simple blinky example for QUTest located in `examples/qutest/blinky`. Also, this release fixes the QF_INT_DISABLE() / QF_INT_ENABLE() macros in the QUTest ports to use the QP namespace. Without this fix, the code fails to compile when the directive "using namespace QP" is not used.
Finally, this release adjusts the code in qs_rx.cpp to use QF_EVT_REF_CTR_INC_() instead of accessing the event's reference counter directly. This direct access fails in case, when event constructors are configured (Q_EVT_CTOR is defined). In that case, the QP::QEvt class protects the reference counter, so it cannot be accessed directly. This option is currently used only in the QP/C++ port to Qt.
------------------------------------------------------------------------------
@section qpcpp_6_3_0 Version 6.3.0, 2018-05-10
The main purpose of this release is fixing the sub-machine support in the QP/Spy build configuration. Specifically, this release fixes the following bug:
- [bug#213 "QP/C/C++ applications with submachines occasionally crash in Spy build configuration"](https://sourceforge.net/p/qpc/bugs/213/)
@attention
This release matches [QM 4.2.0](https://www.state-machine.com/qm/history.html#qm_4_2_0).
Additionally, the release contains some re-factoring of the QS-RX input channel.
------------------------------------------------------------------------------
@section qpcpp_6_2_2 Version 6.2.0, 2018-03-16
The main purpose of this release is extednding the functionality of the [QUTest unit testing](https://www.state-machine.com/qtools/qutest.html) for QP/C++ applications. Specifically, this release adds support for testing of **self-posting** of events in active objects, which is an essential element in the [Reminder](https://www.state-machine.com/doc/Pattern_Reminder.pdf) and [Deferred Event](https://www.state-machine.com/doc/Pattern_DeferredEvent.pdf) design patterns. To implement this new feature, the QS-RX (QS receive channel) has been extened by a small scheduler that processes all secondary events gnenerated by dispatching, posting, or publishing events (only active when the #Q_UTEST macro is defined). Also, the implementation of the target resident QUTest components (file `src/qs/qutest.cpp` and `src/qf/qf_actq.cpp`) have been modified to allow posting of events during unit testing.
Additionally, the release adds standard QS trace records (the `qs.h` header file) for event deferring and recalling (::QS_QF_ACTIVE_DEFER and ::QS_QF_ACTIVE_RECALL) as well as recall-attempt (::QS_QF_ACTIVE_RECALL_ATTEMPT). Also standard QS trace records have been added for creating/deleting new event references (::QS_QF_NEW_REF and ::QS_QF_DELETE_REF, respectively). To make room for these new records, the following rarely-used records have been removed: QS_QF_ACTIVE_ADD, QS_QF_ACTIVE_REMOVE, QS_QF_EQUEUE_INIT, and QS_QF_MPOOL_INIT. The global filter settings in the QS_filterOn() and QS_filterOff() functions have been updated to the augmented and re-organized QS trace records.
@note
Because of the changes in the standard QS trace recods, this release requires the matching **QSPY 6.2.0** or later. Also, because of the changes this release might break some existing QUTest unit test scripts, which need to be re-adjusted to the new trace records.
Additionally, this release improves support for tracing and unit-testing embedded POSIX targets, such as embedded Linux, or OSes/RTOSes with the POSIX compatiblity layer. Specifically, the POSIX ports (`qpcpp/ports/posix/` and `qpcpp/ports/posix-qutest`) no longer produce libraries. Instead all examples for POSIX (`qpcpp/examples/posix/` and `qpcpp/ports/qutest`) build the QP/C framework directly from the sources, which promotes consisentcy in the toolchain and options used. For unit testing, `posix.mak` makefiles have been added for projects `qpcpp/examples/qutest/dpp` and `qpcpp/examples/qutest/self_test`. These makefiles support remote testing of embedded POSIX targets over TCP/IP, where the POSIX target runs only the test fixture, but the host (e.g., Windows host) executes both the QSPY host application and runs the QUTest scripts. The previous makefiles for POSIX have been renamed to `posix_host.mak`, because they are intended to use POSIX as a *host*, as opposed to a *target*.
Also, this release adds new QUTest exampes, which illustrate:
- event deferral (`qpcpp/examples/qutest/defer`)
- dispatching/posting events with parameters (`qpcpp/examples/qutest/evt_par`).
Additionally, this release updates the QP/C++ ports to win32-qv and posix-qv to allow a "tickless" mode, where the "tickerThread" is not created. This mode is set by configuring the system clock tick rate to 0 (QF_setTickRate(0)).
Finally, this release phases out the `qp_port.h` header file. If any of your projects still includes this file, please replace it with the qpc.h heder file.
------------------------------------------------------------------------------
@section qpcpp_6_1_1 Version 6.1.1, 2018-02-18
The main purpose of this release is adding the context-switch callbacks to the preemptive @ref qk "QK" and @ref qxk "QXK" kernels. The QK_onContextSw() and QXK_onContextSw() callback functions provide a mechanism to perform additional custom operations when QK/QXK switches context from one thread to another. To avoid extra overhead when this functionality is not needed and for backwards-compatiblity with the existing applications, the callbacks are enabled only when the macros #QK_ON_CONTEXT_SW (for QK) and #QXK_ON_CONTEXT_SW (for QXK) are defined. These macros can be defined either directly in command-line for the compiler, or in the QK/QXK port files (qk_port.c for QK and qxk_port.c for QXK). Examples for the context-switch callbacks have been provided for for the @ref arm-cm_dpp_nucleo-l053r8 "NUCLEO-L053R8" (Cortex-M0+) and the @ref arm-cm_dpp_nucleo-h743zi "NUCLEO-H743ZI" (Cortex-M7).
Also, this release changes the ARM Cortex-M ports for the IAR-ARM toolchain in that it replaces the assembly modules with the equivalent C implementation. This change enables using the configuration macros #QK_ON_CONTEXT_SW (for QK) and QXK_ON_CONTEXT_SW (for QXK) in the ports. All existing example projects for IAR-ARM have been updated to use the `q(x)k_port.c` files instead of `q(x)k_port.s` files.
Also, this release adds new project files for the Atollic TRUEstudio for STM32. The TRUEstudio projects (Eclipse) have been added for the @ref arm-cm_dpp_nucleo-l053r8 "NUCLEO-L053R8" (Cortex-M0+) and the @ref arm-cm_dpp_nucleo-h743zi "NUCLEO-H743ZI" (Cortex-M7).
Also, this relese updates the CMSIS to version 5.3.0 (see `qpcpp/3rd_party/CMSIS`
Finally, this release fixes the following bug:
- [bug#203 QXK: precondition in QXSemaphore::wait() incorrectly handles the scheduler lock"](https://sourceforge.net/p/qpc/bugs/203/)
------------------------------------------------------------------------------
@section qpcpp_6_1_0 Version 6.1.0, 2018-02-04
The main purpose of this release is adding the support for the [<strong>ARM Compiler 6 (ARM-Clang)</strong>](https://developer.arm.com/products/software-development-tools/compilers/arm-compiler), which is a C/C++ toolchain for ARM processors based on the modern [Clang frontend](https://clang.llvm.org/) and the [LLVM framework](https://llvm.org/). This release adds the ARM-Clang ports and examples for all built-in kernels (QV, QK, and QXK) for the ARM Cortex-M CPU cores.
This release also adds support for the STM32H7 high-performance Cortex-M7 with the **double-precision FPU** (FPv5-DP-D16-M). Specifically, this release provides examples for the [NUCLEO-H743ZI board](http://www.st.com/en/evaluation-tools/nucleo-h743zi.html) (Cortex-M7 with FPv5-DP-D16-M). The @ref arm-cm_dpp_nucleo-h743zi "examples for NUCLEO-H743ZI board" include all built-in kernels with ARM-CLANG, ARM-KEIL, GNU-ARM, and IAR-ARM. Additionally the NUCLEO-H743ZI examples also include the QP FreeRTOS with ARM-KEIL, GNU-ARM, and IAR-ARM.
Also, this release changes the ARM Cortex-M ports for the GNU-ARM toolchain in that it replaces the assembly modules with the equivalent C implementation. This change enables using wider range of GNU-ARM toolchain distributions, such as GNU-ARM Linaro, which had trouble with the assembly modules, but compiles correctly the C modules. All related GNU-ARM example projects have been modifed to use the new C-ports.
Also, this release fixes build problems with IAR EWARM caused by the CMSIS v5.1.1, where the IAR C++ compiler failed to compile the CMSIS "mpu_armv7.h" header file. The problem was caused by the `restrict` keyword, which was supported by default in the IAR C compiler, but was **not** supported by default by the IAR C++ compiler. The fix was to add explicit compiler option `--enable-restrict` to all IAR projects for Cortex-M processors.
Also, this release updates the existing projects for the Code Composer Studio (CCS) to CCSv7. All existing CCS projects for the EK-TM4C123GXL (TivaC LauchPad) and LaunchXL2-TMS57012 (Hercules) boards have been upgraded and tested.
Also, all examples for MSP430 with CCS have been updated to CCSv7 and re-tested on the supported boards.
------------------------------------------------------------------------------
@section qpcpp_6_0_4 Version 6.0.4, 2018-01-10
The main purpose of this release is the provision of the official @ref freertos "QP/C++ port to FreeRTOS" (version 10). The QP/C++ port to FreeRTOS is completely generic and should work on any CPU supported by FreeRTOS. The port comes with the following examples (see examples/freertos/):
- @ref freertos_dpp_ek-tm4c123gxl "DPP on EK-TM4C123GXL (ARM Cortex-M4F)" with ARM-KEIL, GNU-ARM and IAR-ARM toolchains.
- @ref freertos_dpp_stm32f746g-disco "DPP on STM32F746G-Discovery (ARM Cortex-M7)" with ARM-KEIL, GNU-ARM and IAR-ARM toolchains.
This release also replaces assembly with C implementation in the ARM-KEIL ports or the QK and QXK kernels to Cortex-M to take advantage of the `__asm` functions. (NOTE: this change has impact on the existing QP/C++ applications that use the ARM-KEIL toolchain and the QK or QXK kernels.)
Additionally, this release modifies all IAR-ARM example projects by adding compiler option "--enable_restrict". This option is necessary to compile the CMSIS header file "mpu_armv7.h" under the IAR C++, where the "restrict" keyword is not supported by default.
Additionally, this release fixes some problems with the native examples for STM32F4 and STM32F7 boards.
Additionally, this release adds a generic function QP::QActive::setAttr() to set thread attributes in various QP ports to 3rd-party RTOSes. This function is then used in the @ref embos "embOS port" and the @ref ucos-ii "uC/OS-II port".
Finally, this release fixes the following bug in the @ref threadx "ThreadX port":
- [bug#197 "Problems with scheduler locking in QP/C/C++ ports to ThreadX"](https://sourceforge.net/p/qpc/bugs/197/)
------------------------------------------------------------------------------
@section qpcpp_6_0_3 Version 6.0.3, 2017-12-12
Changes since version 6.0.1:
- fixed [bug#193 "QXK: context switch from extended to basic thread fails"](https://sourceforge.net/p/qpc/bugs/193/)
- replaced attributes of QP::QActive and QP::QEQueue classes from type `uint_fastX_t` to `uintX_t`. For the ARM Cortex-M port, these changes reduce the memory footprint of each QP::QActive instance from 50 to 38 bytes (25% improvement).
- improved the performance of the ARM Cortex-M0/M0+ ports, which do not supprt the CLZ (count leading zeros) instruction. All QP ports to Cortex-M0/M0+ (qv, qk and qxk) now contain hand-optimized assembly implementation of the LOG2 (log-base-2) function. The implementation executes only in 14 machine instructions and uses only 16-byte look-up-table (LUT). This replaces LOG2 implementation based on a 256-byte LUT.
- changed the interrupt disabling policy in ARM Cortex-M ports back to "unconditional interrupt disabling", which is simpler and faster than "saving-and-restoring interrupt status".
- added a port to ARM Cortex-M with the LLVM compiler (currently only for the dual-mode QXK kernel). Unlike the other ARM Cortex-M ports, the port to LLVM uses the "saving-and-restoring interrupt status" interrupt disabling policy (explicit request from a commercial customer).
- removed ARM Cortex-M port and examples for the TI-ARM compiler. (Projects for the TI CCS can use the GNU-ARM toolchain and the existing GNU-ARM ports).
- Added the DPP-COMP example (Dining Philosophers Problem) with the "Orthogonal Component" state pattern located in `qpcpp/examples/win32/dpp-comp`. This example demonstrates also a partitioned QM model into external packages for the Container (Table active object) and the Components (Philo active objects).
@note
This release does not change any of the QP/C++ APIs.
------------------------------------------------------------------------------
@section qpcpp_6_0_1 Version 6.0.1, 2017-12-10
The main focus of this release is to fix the remaining problems with transitions out of eXit-Points in sub-machines. Specifically, this release modifies the QMsm-based state machine implementation strategy (file src/qf/qep_msm.cpp to properly handle transitions from eXit-Points to Entry-Points and to History connectors in sub-machines. These changes are part of fixing the following bugs reported for QM:
- [bug#190 "Exit-Point segment targeting History doesn't work](https://sourceforge.net/p/qpc/bugs/190/)
- [bug#189 "Exit-Point segment targeting an Entry-Point to sub-machine state doesn't work"](https://sourceforge.net/p/qpc/bugs/189/)
Additionally, this release fixes the following bug in transitions to "shallow history":
- [bug#191 "Transition to shallow history in QMsm causes assertion qep_msm#810"](https://sourceforge.net/p/qpc/bugs/191/)
The bug#191 is fixed by modifying the function QMsm::childStateObj() in qep_msm.cpp to return the parent state in the corner case when the current state is the parent state.
@note
This QP/C++ 6.0.1 release is the minium version required in the upcoming QM 4.1.0. This is because QM 4.1.0 assumes the modified QMsm-state machine implementation strategy in order to properly handle the various transitions out of eXit-Points in sub-machine states.
Additionally, this release changes the QXK implementation still related to the [bug#186 "QXK: Extended thread context switch causes assertion in PendSV_Handler"](https://sourceforge.net/p/qpc/bugs/186/). Specifically, the case of context switching away and back to the same thread (which can arise under specific interrupt preemption scenarios) is now handled as a simple return from PendSV. The QXK scheduler has been modified to set the "next" thread pointer to NULL when it detects switching back to the "current" thread.
------------------------------------------------------------------------------
@section qpcpp_6_0_0 Version 6.0.0, 2017-10-13
This release fixes two bugs found in the QXK kernel:
- [bug#185](https://sourceforge.net/p/qpc/bugs/185/) "QXK: PendSV_Handler uses inconsistent stack frames for saving and restoring AO for Cortex-M0(+)"
- [bug#186](https://sourceforge.net/p/qpc/bugs/186/) "QXK: Extended thread context switch causes assertion in PendSV_Handler"
Additionally, this release includes a workaround for a [bug indetified in the GNU-ARM toolset](https://bugs.launchpad.net/gcc-arm-embedded/+bug/1722849). This problem affected the ARMv6-M architecture (Cortex-M0/M0+/M1) and manifested itself in generation of incorrect code for the QP critical section at certain gcc optimization levels (such as -O). This bug was first discovered and filed as [bug#184](https://sourceforge.net/p/qpc/bugs/184/). The bug affected the GNU-ARM ports to all built-in kernels @ref qv "QV", @ref qk "QK", and @ref qxk "QXK".
@note
This release no longer contains the directory `qpcpp/source`, which was scheduled to be phased out in QP5. In QP6 the source code is found only in the `qpcpp/src` directory.
-----------------------------------------------------------------------------
@section qpcpp_5_9_9 Version 5.9.9, 2017-09-29
This release implements the feature request [#132 "Extend the QXK mutex to support also simple operation without the priority-ceiling protocol"](https://sourceforge.net/p/qpc/feature-requests/132/).
@note
This release is completely backwards-compatible with the @ref qpcpp_5_9_8 "previous QP/C++ release 5.9.8".
Specifically, the QP::QXMutex class has been extended as follows:
- a mutex initialized with ceiling==0 (see QP::QXMutex::init()) means "no ceiling", so such mutex will NOT use the priority-ceiling protocol.
- in this case the mutex will NOT require a unique priority level
- in this case the mutex will not change (boost) the priority of the holding thread in the QP::QXMutex::lock() operation.
- in this case the mutex will still support nesting of locks (as before), up to 255 levels of nesting.
- when initialized with `ceiling>0`, QP::QXMutex WILL use the priority-ceiling protocol, as before. It will require that the ceiling priority be unique and not used by any other thread or mutex. In other words, the previous functionality remains unchanged.
-----------------------------------------------------------------------------
@section qpcpp_5_9_8 Version 5.9.8, 2017-09-15
This release fixes the QXK kernel [bug#182 " Inconsistent QXThread::post_() behavior with respect to the 'margin' parameter"](https://sourceforge.net/p/qpc/bugs/182/).
Also, the pre-condition assertion in the function QF::newRef_() (file `src/qf/qf_dyn.cpp`) has been modified to allow creating event references only for dynamic events `(e->poolId_ == 0)`.
Modifed ARM Cortex-M examples with the GNU-ARM toolset to be compatible with C++ RTTI and C++ Exception handling, as follows:
- removed definitions of `_init()` and `_fini()` from the GNU-ARM startup code for all supported boards in the `3rd_party` directory
- added the empty definitions of `_init()` and `_fini()` to `mini_cpp.cpp` in the `3rd_party/gnu_cpp directory`
- Modified all GNU-ARM linker scripts (.ld files) to add the following symbols:
@verbatim
__exidx_start = .;
.ARM.exidx : { *(.ARM.exidx* .gnu.linkonce.armexidx.*) } >RAM
__exidx_end = .;
@endverbatim
Provided C++ example with RTTI and Exception hanlding in the directory:
`examples/arm-cm/dpp_efm32-slstk3401a/qxk/gnu_with-rtti`
-----------------------------------------------------------------------------
@section qpcpp_5_9_7 Version 5.9.7, 2017-08-18
The main focus of this release are new requested features for the @ref qxk "dual-mode QXK kernel":
- <a href="https://sourceforge.net/p/qpc/feature-requests/129" class="extern">feature#129 "Allow blocking while holding a mutex in QXK"</a>; and
- <a href="https://sourceforge.net/p/qpc/feature-requests/130" class="extern">feature#130 "Allow QXK mutex locks to nest while acquired by the same thread"</a>; and
- <a href="https://sourceforge.net/p/qpc/feature-requests/131" class="extern">feature#131 "Add a non-blocking "tryLock()" operation to the QXK mutex"</a>.
Additionally, this release adds also the non-blocking QXSemaphore::tryWait() operation.
In the process of re-implementing the QP::QXMutex, the previouis non-blocking priority ceiling mutex has been replaced with an equivalent selective QXK scheduler locking up to the specified ceiling priority. Specifically, this feature has been implemented with two new operations QXK::schedLock() and QXK::schedUnlock().
For consistency, the non-blocking mutex of the @ref qk "QK kernel" has been also replaced by the operations QK::schedLock() and QK::schedUnlock().
All related QXK and QK examples have been updated to use the selective scheduler locking instead of the mutex. The new blocking QXK mutex has been demonstrated in the following updated examples:
- @ref arm-cm_dpp_efm32-slstk3401a
- @ref arm-cm_dpp_ek-tm4c123gxl
@attention
The changes to QP::QXMutex in QXK and the now obsolete QMutex in QK **break backwards-compatiblity** with the exising code that relies these features. In the exising code, the mutexes should be replaced with selective scheduler locking.
@attention
Also, the blocking APIs with timeouts, such as QXMutex::lock(), QXSemaphore::wait(), QXThread::delay(), and QXThread::queueGet() no longer require the last `tickRate` parameter. This also **breaks backward-compatiblity** with the exising code that uses these operations.
This release also updates CMSIS to version 5.1.0 (3rd_party/CMSIS).
This release also adds the @ref api "API Reference" section to the QP/C++ documentation.
-----------------------------------------------------------------------------
@section qpcpp_5_9_6 Version 5.9.6, 2017-08-04
The main focus of this release are improvements to the "dual-mode" QXK kernel. Specifically, this release implements the <a href="https://sourceforge.net/p/qpc/feature-requests/128/" class="extern">featrue request #128 "QP Semaphore Max Value Setting"</a> for QXK. This feature changes the QXK function QP::QXSemaphore::init(), which now takes additional parameter `count`. This parameter specifies the maximum allowed count for the semaphore (e.g., count of 1 makes the semaphore a binary-semaphore).
@note
This change does NOT break any existing code, because the additional parameter has a default value (0xFFFF).
This release also fixes a bug inside the assertions in QP::QXSemaphore::signal(), where the check for the extended-thread is performed.
Also, this release adds assertions to the QXK code, which ensure that any available blocking calls can only be made from extended-threads and not from basic-threads (active objects) or ISRs.
Also, this release adds protection in the IRQ priorities initialization in QK/QV/QXK for ARM Cortex-M3/M4/M7, so that the number of IRQs is extracted from bits 0-2 of the ICTR register (INTLINESNUM).
Finally, this release consistently changes all example projects (for all toolchains) to use the **src/** directory for QP/C++ source code, instead of the **source/** directory. The source/ directory is now truly obsolete, but is still provided in this release for backwards compatibility with user projects.
-----------------------------------------------------------------------------
@section qpcpp_5_9_5 Version 5.9.5, 2017-07-20
This release fixes the following bugs:
- <a href="https://sourceforge.net/p/qpc/bugs/178/" class="extern">bug#178</a> "GNU-ARM compiler reports "Error: unaligned opcodes..." in startup code for QP/C/C++/nano examples". The bug fix entails modifying the startup code for the GNU-ARM compiler in the `3rd_party` directory. Specifically, the proper alignment directives have been added to the inline assembly in the exception handlers.
- <a href="https://sourceforge.net/p/qpc/bugs/179/" class="extern">bug#179</a> "Assertion ID 210 fires when signaling on a QXK semaphore"
This release fixes the naming problem of the startup code for the
STM32F7-Discovery board (in the `3rd_party/stm32f7-discovery/gnu/ and arm/`
directories), where the startup code was renamed from `startup_stm32f4xx.c/s`
to `startup_stm32f746xx.c/s`. The change has been also made in the example
projects for the STM32F7-Discovery board (for ARM-KEIL and GNU-ARM toolsets).
Finally, this release modifies the example code (mostly `bsp.cpp` files)
so that the Q_DEFINE_THIS_FILE() macro is used consistently *outside*
a namespace at the top of the file scope.
-----------------------------------------------------------------------------
@section qpcpp_5_9_4 Version 5.9.4, 2017-07-07
This release adds Thread-Local Storage (TLS) feature for the dual-mode
QXK kernel (see @ref qxk_tls).
-----------------------------------------------------------------------------
@section qpcpp_5_9_3 Version 5.9.3, 2017-06-19
This release implements the feature request
<a href="https://sourceforge.net/p/qpc/feature-requests/126/" target="_blank" class="extern">#126</a> "Allow non-asserting
event allocation for zero-margin allocations". Specifically, calling
Q_NEW_X() or QP::QActive::post_() with the @p margin argument of zero will no
longer assert if the allocation/posting fails.
-----------------------------------------------------------------------------
@section qpcpp_5_9_2 Version 5.9.2, 2017-06-05
This release adapts the Makefiles for GNU-ARM to the new location of the
GNU-ARM toolset, which is now included in the QTools Collection (v 5.9.1)
for Windows.
Also, this release improves the flash loading scripts for the JLink
hardware debugger (for GNU-ARM projects for the EFM32-SLSTK3401A board).
Specifically, the JLink configuration file for flash download is
generated by the flash batch script based on the command-line paramter
(the binary file to load into the flash). This eliminates the need to
manually maintain JLink configuration files.
Also, this release adds bi-directional QP/Spy to the embOS example
project for the STM32F4-Discovery board.
Also, this release adds GNU-ARM port to uC/OS-II and adds GNU-ARM example
project for the EK-TM4C123GXL board.
Finally, this release implements the feature request #125 "Include QPC
Demo application for STM32F4 processor without RTOS" (see
https://sourceforge.net/p/qpc/feature-requests/125/ ). The DPP demo for
the STM32F4-Discovery board has been added in the directory:
qpc/examples/arm-cm/dpp_stm32f4-discovery . This demo includes QV, QK and
QXK kernels and ARM-Keil, GNU-ARM, and IAR-ARM toolsets. The demos support
bi-directional QP/Spy.
------------------------------------------------------------------------------
@section qpcpp_5_9_1 Version 5.9.1, 2017-05-26
This release fixes the following bug:
- <a href="https://sourceforge.net/p/qpc/bugs/169/" target="_blank" class="extern">bug#169 "Submachine-state eXit Point does not work correctly"</a>
@note
The bug only affects sub-machines and does not affect any other
aspects of QMsm-style state machines.
Also, this release changes the organization of the QP/C++ source code to
make it more friendly for the Eclipse CDT, as proposed in the <a href="https://sourceforge.net/p/qpc/feature-requests/123/" target="_blank" class="extern">feature
request #123 "Eclipse-friendly source directory structure for QP/C/C++"</a>.
Specifically, the QP/C++ source code is now provided in the
`qpc/src/` directory with the following structure:
@verbatim
qpcpp/
+-source/ - exitsting source directory with "flat" structure
| (for backwards-compatibility)
+-src/ - new source directory grouped by functionality
+-qf/ - core framework (QEP + QF)
+-qk/ - QK kernel
+-qv/ - QV kernel (only one file qv.c)
+-qxk/ - QXK kernel
+-qs/ - QS software tracing
+-qf_pkg.h
+-qs_pkg.h
+-qxk_pkg.h
@endverbatim
@note The original `qpcpp/source` directory is still provided for backwards
compatibility with the existing QP/C projects. This directory will be
phased out in the future QP/C++ releases. Please use the new source code
structure provided in the `qpcpp/src` directory.
------------------------------------------------------------------------------
@section qpcpp_5_9_0 Version 5.9.0, 2017-05-19
The main purpose of this milestone QP/C++ release is to provide support for the powerful **Unit Testing Framework** called
<a href="https://state-machine.com/qtools/qutest.html" target="_blank" class="extern"><strong>QUTest&trade;</strong> (pronounced *cutest*)</a>.
QUTest&trade; is the fundamental tooling for Test-Driven Development (TDD) of QP/C++ applications, which is a highly recommended best-practice. This release introduces changes in the QS-RX (receive) channel and adds several new callbacks.
@note The signature of the QS::onCommand() has changed and the function now takes 3 arbitrary 32-bit parameters instead of one. This introduces backwards-incompatibility with previous code that used QS::onCommand().
This release also changes the critical section for QP/C++ ports to ARM Cortex-M
in that the policy of "save and restore interrupt status" is used. This policy
permits nesting of critical sections, which was requested by customers.
Additionally, this release changes the selective interrupt disabling for
ARM Cortex-M3/4/7 (with the BASEPRI register) to address the hardware problem
on ARM Cortex-M7 core r0p1 (ARM-EPM-064408, errata 837070). The QP ports to
ARM Cortex-M3/4/7 now implement the workaround recommended by ARM, which is
to surround MSR BASEPRI with the "CPSID i"/"CPSIE i" pair. This workaround
works also for Cortex-M3/M4 cores.
New ports:
- ports/win32-qutest folder contains port to QUTest for Windows
- ports/posix-qutest folder contains port to QUTest for POSIX (Linux)
- ports/arm-cm/qutest folder contains port to QUTest for ARM Cortex-M
New examples:
- examples/qutest/dpp folder contains the QUTest DPP test for various
platforms (Win32, EFM32-SLSTK3401A and EK-TM4C123GXL)
- examples/qutest/qhsmtst folder contains the QUTest test for
the QHsmTst state machine (structural test)
- examples/qutest/qmsmtst folder contains the QUTest test for
the QMsmTst state machine (structural test)
- examples/qutest/self_test folder contains the QUTest self-test for
various features of QUTest
Updates of 3rd_party software:
- the 3rd_party/CMSIS folder has been updated to CMSIS 5.0.2.
- the 3rd_party/embOS folder has been updated to embOS 4.34.1
Finally, this release fixes the following bugs:
- bug#162 "QF critical sections require modification for M7 core"
------------------------------------------------------------------------------
@section qpcpp_5_8_2 Version 5.8.2, 2017-02-08
This release adds examples for the ARM Cortex-<strong>M7</strong> CPU. Specifically,
the release contains the standard @ref dpp "Dining Philosophers Problem (DPP)"
examples for the @ref arm-cm_dpp_stm32f746g-disco "STM32F746G-Discovery board",
all build-in kernels (QV, QK, and QXK), and ARM-KEIL, IAR EWARM, GNU-ARM toolsets.
To provide examples for STM32F746G-Discovery board, the release now
provides the folder `3rd_party/stm32f7-discovery` with the support code
for the STM32F7xx MCUs, which contains parts of STM32CubeF7 library.
Also, the `3rd_party/CMSIS` folder now provides the new CMSIS V5.0.1.
Also, the file `3rd_party/gnu_cpp/mini_cpp.cpp` has been updated to provide
compatibility with the latest GNU-ARM (gcc 6.2.0).
Finally, this release fixes the following bugs:
- bug#159 QP/C/C++ Win32 ports don't work on all x86 CPUs
- bug#157 In QPC ucosii port, conversion of AO's priority to OS task
priority is incorrect.
- bug#152 Typo (qpcpp/ports/arm7-9/qk/gnu/qk_port.s:42) prevents
compilation
- bug#151 QTicker compile error when #Q_EVT_CTOR is defined
- bug#154 qspy.c parses history transitions incorrectly
------------------------------------------------------------------------------
@section qpcpp_5_8_1 Version 5.8.1, 2016-12-16
This release is in response to a recent finding that many QP users of the
ports to ARM Cortex-M3/M4 forget to explicitly set their interrupt priorities,
as described in the AppNote
<a href="https://www.state-machine.com/doc/AN_ARM-Cortex-M_Interrupt-Priorities.pdf"
target="_blank" class="extern">"Setting ARM Cortex-M Interrupt Priorities in QP 5.x"</a>.
Specifically, this release improves safety of QP ports to ARM Cortex-M3/M4,
by initializing the interrupt priorities to a safe default in a generic,
portable way. This QP port includes such a fix for QV/QK/QXK ports to ARM Cortex-M3/M4.
Additionally, this release introduces the new QP::QTicker class, which is an
efficient active object specialized to process QF system clock tick at a
specified tick frequency [0..#QF_MAX_TICK_RATE]. Placing system clock tick
processing in an active object allows you to remove the non-deterministic
TICK_X() processing from the interrupt level and move it into the thread-level,
where you can prioritize it as low as you wish.
Changes in detail:
- modified the QV, QK, and QXK source code to call QV_init(), QK_init(), and
QXK_init(), respectively.
- modified the ARM Cortex-M ports of QV, QK, and QXK to initialize
priorities all exceptions and IRQs to the safe value #QF_BASEPRI.
@note The QV port now has a new **qv_port.cpp** module that needs to
be added to the build.
- added declaration of the QP::QTicker class to qf.h
- added implementation of the QP::QTicker class to qf_actq.cpp
- modified the following examples to demonstrate the use of the QP::QTicker:
+ qpcpp/examples/arm-cm/dpp_efm32-slstk3401a/qk
+ qpcpp/examples/arm-cm/game_efm32-slstk3401a/qv
+ qpcpp/examples/arm-cm/game_efm32-slstk3401a/qk
- added the header file **cmsis_ccs.h** to qpcpp/3rd_party/CMSIS/Include directory
(used in the examples for the Code Composer Studio). The file has been
dropped during the upgrade to CMSIS 5.0.1, because it is not part of the
standard distribution.
------------------------------------------------------------------------------
@section qpcpp_5_8_0 Version 5.8.0, 2016-11-30
The main purpose of this milestone QP/C++ release is to finally provide
the baseline framework fully compatible with the upcoming QM 4.0.0.
This release changes the class hierarchy so that QP::QHsm becomes the base
class of QP::QMsm and QP::QActive. Also, QP::QActive becomes the base class of
QP::QMActive, which reverses the changes introduced in version 4.1.
The modified @ref classes "class hierarchy" better reflects the fact that QP::QHsm state
machine implementation strategy is simpler and supports less functionality
than the more advanced QP::QMsm strategy. For example, only the QP::QMsm class
fully supports sub-machines and sub-machine states that are the main
feature of QM 4.x. This clean progression of supported functionality from
subclasses to superclasses allows QM to easier check and enforce that
advanced features are not generated for subclasses that don't have the
required capabilities. (With previous class hierarchy with QP::QMsm as the
base class all subclasses, including QP::QHsm, would technically inherit the
advanced functionality, which is not the case).
@note
The changes in QP/C++ 5.8.0 might occasionally require changing existing
QP/C++ applications by changing the type of pointers to QP::QMsm with
pointers to QP::QHsm or pointers to QP::QMActive with QP::QActive,
because of the reverse superclass-subclass relationship between these
classes.
Also, this release changes the implementation of the QV, QK, and QXK kernels
in that the ready-set representing active threads is cleared only *after*
completion of the RTC-step, not when the last event is removed from the
corresponding event queue. In case of the QXK kernel this change fixes the
high-priority <a href="https://sourceforge.net/p/qpc/bugs/147/">bug#147</a>.
But even in case of the QV and QK kernels, where this behavior didn't lead
to any bugs, the policy better reflects the semantics of the ready-set.
This release also updates the CMSIS interface included in the 3rd_party/CMSIS
folder to the latest CMSIS-5.
All examples and QM models have been updated to the new upcoming QM 4.0.0.
All these models require QM 4.x.
Finally, the complete list of bugs fixed in this release is as follows:
- <a href="https://sourceforge.net/p/qpc/bugs/147/">bug#147</a> "QXK: PendSV_error is triggered on special conditions"
- <a href="https://sourceforge.net/p/qpc/bugs/144/">bug#144</a> "Obsolete Win32 API in qwin_gui.c"
- <a href="https://sourceforge.net/p/qpc/bugs/124/">bug#124</a> "Windows port now cause memory leakage"
------------------------------------------------------------------------------
@section qpcpp_5_7_4 Version 5.7.4, 2016-11-04
This release fixes the following bugs:
- bug#145 QP::QF::publish_() leaks events that have no subscribers
- bug#144 Obsolete Win32 API in qwin_gui.c
- bug#143 QP::QMActive::postLIFO on initial transition asserts on QXK
------------------------------------------------------------------------------
@section qpcpp_5_7_3 Version 5.7.3, 2016-10-07
This release adds QP ports to the TI-RTOS kernel (SYS/BIOS) with TI-CCS
and IAR EWARM toolsets. Examples are provided for the EK-TM4C123GXL
(TivaC LaunchPad) in the directory:
qpcpp/examples/ti-rtos/arm-cm/dpp_ek-tm4c123gxl
NOTE: The examples require a separate installation of the TI-RTOS
(file tirtos_tivac_setupwin32_2_16_01_14.exe)
Also, this release fixes the following bugs:
- bug#140 (PendSV_Handler() exception stacked PC not halfword aligned).
- bug#142 (PendSV_restore_ex may not be able to enable interrupt before
returning to task).
------------------------------------------------------------------------------
@section qpcpp_5_7_2 Version 5.7.2, 2016-09-30
This is the first production release of the "dual-mode" QXK kernel.
"Dual-mode" QXK means that QXK supports both basic-threads (BC1 class
from the OSEK/VDX RTOS specification) as well as extended-threads
(EC1 class from the OSEK/VDX RTOS specification. In other words,
QXK executes active objects (basic threads) like the QK kernel using the
single stack (Main Stack on ARM Cortex-M), but can also execute
traditional *blocking* threads (extended threads).
Only the extended threads (QP::QXThread class) need their private stack
spaces and the overhead of the full context switch. The basic threads
(QP::QMActive and QP::QActive classes) run efficiently using the main stack with
much lower context switch overhead.
The QXK examples have been updated for more thorough demonstration of
the QXK features. The QXK examples are available in the following
directories: dpp_efm32-slstk3401a, dpp_ek-tm4c123gxl, and dpp_nucleo-l053r8.
This release also changes the internal QK implementation to match the
terminology applied in the QXK kernel (e.g., QK_sched_() has been renamed
to QK_activate_() and QK_schedPrio_() to QK_sched_()). These changes fall
into the category of refactoring and have no impact on the API or
performance.
Finally, this release improves the implementation of scheduler locking
in publish-subscribe event delivery.
------------------------------------------------------------------------------
@section qpcpp_5_7_0 Version 5.7.0, 2016-08-31
This release adds support for sub-machines and sub-machine states for reusing pieces of state machines (an advanced UML concept) to the QMsm-state machine implementation strategy. This feature is to match the upcoming QM 4.0.0.
Also, this release adds support for the ARM Cortex-R processor. Specifically, the release contains a generic port to ARM Cortex-R with the IAR and TI-CCS toolsets and @ref arm-cr_dpp_launchxl2-tms57012 "examples for the TI Hercules TMS570LS12x" safety MCU (LAUNCHPADXL2-TMS57012).
Also, this release changes once more the QK port to ARM Cortex-M, to reduce the interrupt latecy. This has been achieved by shortening the critical section in the PendSV exception. Also, this QK port to Cortex-M no longer uses the NMI exception.
Also, this release changes slightly the QXK port to ARM Cortex-M, where again the critical section in PendSV has been slighly shortened.
Finally, this release replaces all absolute paths with relative paths in all CCS-Eclipse project files (for TivaC, Hercules, and MSP430).
Changes in detail:
1. Modified qep_msm.cpp to correct the support for sub-machines and sub-machine states
2. Added new port to ARM Cortex-R in the directory ports/arm-cr
3. Added examples for ARM Cortex-R in the directory examples/arm-cr
4. Modified the ARM Cortex-M QK ports (ARM-KEIL, GNU, IAR, and TI)
5. Modified the ARM Cortex-M QXK ports (ARM-KEIL, GNU, IAR, and TI)
------------------------------------------------------------------------------
@section qpcpp_5_6_5 Version 5.6.5, 2016-06-06
This release adds support for the new board: EFM32-SLSTK3401A (Pearl Gecko Starter Kit from Silicon Labs). This board replaces the Stellaris EK-LM3S811 board, which has been discontinued. (The Stellaris EK-LM3S811 board had been used in the "Fly 'n' Shoot" game example accompanying the PSiCC2 book).
This release also introduces a new version of the QWIN GUI Toolkit in the Windows prototypes for the "Fly 'n' Shoot" game and the DPP-GUI
version (see https://www.state-machine.com/products/index#QWIN).
Additionally, this release also includes the QP/C integration with the emWin emgedded GUI from SEGGER, which is also the same product as
uC/GUI distributed by Micrium (@ref exa_emwin).
Finally, this relase comes with updated project files for TI Code Composer Studio (both for ARM Cortex-M and for MSP430).
This release fixes the following bugs:
- bug#130 (POSIX port stop->start leads to reuse of sestroyed mutex).
- bug#131 (QP::QF::newRef_() increments reference counter without QF_CRIT_ENTRY_()).
------------------------------------------------------------------------------
@section qpcpp_5_6_4 Version 5.6.4, 2016-04-25
This release fixes a serious Bug #128 (https://sourceforge.net/p/qpc/bugs/128 ) in the QK port to ARM Cortex-M introduced back in QP 5.6.1
------------------------------------------------------------------------------
@section qpcpp_5_6_3 Version 5.6.3, 2016-04-12
This release fixes a serious Bug #127 (https://sourceforge.net/p/qpc/bugs/127 ) in the QK preemptive scheduler introduced in QP 5.6.2.
------------------------------------------------------------------------------
@section qpcpp_5_6_2 Version 5.6.2, 2016-03-31
The main purpose of this release is to introduce _atomic event multicasting_, meaning that event publishing to all subscribers is now protected from preemption. This eliminates potential for re-ordering of events under preemptive kernels (such as QK, QXK, or 3rd-party RTOSes), when events are published from low-priority AOs and some higher-priority subscribers can preempt multicasting and post/publish events of their own (before the original event is posted to all subscribers).
The atomic event multicasting is implemented by means of selective scheduler locking--very much like a priory-ceiling mutex. During event multicasting the scheduler gets locked, but only up to the highest-priority subscriber to a given event. The whole point here is that active objects with priorities above such "priority ceiling" are _not_ affected. Please see the discussion thread:
https://sourceforge.net/p/qpc/discussion/668726/thread/c186bf45
This release also changes the implementation of the priority-ceiling mutex in the preemptive built-in kernels: QK and QXK. Specifically, the implementation now re-uses the selective scheduler locking mechanism. In this new implementation, the QXMutex of the QXK kernel is much more efficient and lightweight, but it _cannot block_ while holding a mutex.
Finally, this release changes the QP ports to 3rd-party RTOSes by performing any RTOS operations (like posting events to message queues) outside critical sections. Also the ports have been augmented to support scheduler locking (this feature depends on what's available in the specific RTOSes).
Changes in detail:
1. Added scheduler locking to QF::publish_() in qf_ps.cpp. This feature is added in a portable way, via macros #QF_SCHED_STAT_TYPE_, QF_SCHED_LOCK_() and QF_SCHED_UNLOCK_(), which need to be implemented in every QP port.
2. Modified QV kernel to provide (dummy) implementation of selective scheduler locking.
3. Modified QK kernel to implement selective scheduler locking via modified priority-ceiling mutex QMutex.
4. Modified QXK kernel to implement selective scheduler locking via modified priority-ceiling mutex QXMutex.
5. Modified embOS port to provide (global) scheduler locking, which affects all priorities, because that's all embOS supports. Also, modified the embOS port to perform event posting outside the QF critical section.
6. Modified uC/OS-II port to provide (global) scheduler locking, which affects all priorities, because that's all uC/OS-II supports. Also, modified the uC/OS-II port to perform event posting outside the QF critical section.
7. Modified ThreadX port to provide selective scheduler locking, by means of "priority-threshold" available in ThreadX. Also, modified the ThreadX port to perform event posting outside the QF critical section.
8. Changed the ThreadX example to run on ARM Cortex-M4 board (STM32DiscoveryF4), instead of Win32 emulation (see qpcpp/examples/threadx/arm-cm/dpp_stm32f429-discovery).
9. Modified the Win32 port to provide (global) scheduler locking, which is implemented by Win32 critical section.
10. Fixed Bug#122 (QP didn't initiate some internal variables) https://sourceforge.net/p/qpc/bugs/122/ by adding explicit clearing of all QP variables in QF::init().
11. Modified the POSIX port to dummy-out scheduler locking. This means that this port currently does NOT lock scheduler around event publishing. (At this point it is not clear how to implement POSIX scheduler locking in a portable way.)
12. Modified QK and QXK examples in qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl to demonstrate the usage of the new priority-ceiling mutexes.
13. Fixed the 3rd-party file startup_stm32l32l1xx.c to include exceptions for Cortex-M3 (MemManage_Handler, BusFault_Handler, and UsageFault_Handler).
14. Updated the 3rd-party files for the EK-TM4C123GXL board (TivaC LaunchPad).
15. Modified Makefiles for the EK-TM4C123GXL board with GNU-ARM toolset to define the symbol TARGET_IS_TM4C123_RB1 for compatibility with the updated 3rd-party files.
16. Implemented Feature Request #110 as well as the duplicate Request #62 by adding function QMActive::flushDeferred()
------------------------------------------------------------------------------
@section qpcpp_5_6_1 Version 5.6.1, 2016-01-01
This release is the first official (production) release of the new blocking @ref qxk "QXK kernel" ("eXtended Quantum Kernel"). QXK is a small, preemptive, priority-based, **blocking** kernel that provides most features you might expect of a traditional blocking RTOS kernel.
QXK has been designed specifically for applications that need to mix event-driven active objects with traditional blocking code, such as commercial middleware (TCP/IP stacks, UDP stacks, embedded file systems, etc.) or legacy software. The QXK kernel is integrated tightly and optimally with the rest of the QP. It reuses all mechanisms already provided in QP, thus avoiding any code duplication, inefficient layers of indirection, and additional licensing costs, which are inevitable when using 3rd-party RTOS kernels to run QP/C++ applications.
@note
The QXK documentation is available in the QP/C Reference Manual at @ref qxk
Additionally, this release removes the macros Q_ROM, Q_ROM_BYTE, and Q_ROM_VAR from the QP/C++ code. These macros have been necessary for odd Harvard-architecture 8-bit CPUs (such as AVR, 8051) to place constant data in ROM. As QP/C++ stopped supporting those CPUs, the non-standard extensions could be removed from the QP/C++ code base.
Additionally, this release re-designs the priority-ceiling mutex in the QK kernel, which now works the same as the mutex of the new QXK kernel. Also, the QK ports to ARM Cortex-M no longer need or use the SVC_Handler (Supervisor Call). This is done to make the QK ports compatible with various "hypervisors" (such as mbed uVisor or Nordic SoftDevice), which use the SVC exception.
Finally, this release modifies the GNU-ARM ports of QK for ARM Cortex-M, to use the __ARM_ARCH macro to distinguish among different architectures (ARCHv6 vs ARCHv7).
Changes in detail:
1. Added new header files for QXK: qxk.h, and qxthread.h.
2. Added new source files for QXK: qxk.cpp, qxk_mutex.cpp, qxk_pkg.h, qxk_sema.cpp, qxk_xthr.cpp.
3. Added QXK ports to ARM Cortex-M for ARM-KEIL, GNU-ARM, IAR, and TI-ARM toolsets (see qpcpp/ports/arm-cm/qxk)
4. Added QXK examples for ARM Cortex-M (in qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl/qxk and qpcpp/examples/arm-cm/dpp_nucleo-l053r8/qxk) for all supported toolsets.
5. Removed Q_ROM, Q_ROM_BYTE, and Q_ROM_VAR from the QP/C++ code.
6. Added Q_ROM, Q_ROM_BYTE to the compatibility-layer in qpcpp.h.
7. Removed ports and examples for the following 3rd-party RTOSes: CMSIS-RTX and FreeRTOS, as QXK provided all the features found in those kernels and is recommended over those kernels.
8. Removed AVR ports and examples.
9. Re-designed the QK priority-mutex in files qk.h and qk_mutex.cpp.
10. Provided QK mutex examples in qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl/qk and qpcpp/examples/arm-cm/dpp_nucleo-l053r8/qk.
11. Updated Makefiles for GNU-ARM to use the __ARM_ARCH macro for defining the ARM architecture.
12. Updated CMSIS from 4.2 to 4.3 in qpcpp/3rd-party/CMSIS
13. Added error directives to source files from different built-in kernels (QV, QK, and QXK) to generate meaningful error messages when these files are mixed in one project. For example, a project based on QK will report errors when source files for QV or QXK are included in it.
14. Corrected example projects for the ARM Cortex-M with TI/CCS toolset
------------------------------------------------------------------------------
@section qpcpp_5_5_1 Version 5.5.1, 2015-10-05
The purpose of this release is to improve the AAPCS compliance of the ARM Cortex-M port to the QK preemptive kernel. Specifically, the PendSV handler in assembly did not always maintain the 8-byte stack alignment, which is required by AAPCS. This version corrects the stack misalignment in the qk_port.s files for all supported ARM compilers (ARM-Keil, GNU, and IAR). All these ports should also be ready for ARM Cortex-M7.
------------------------------------------------------------------------------
@section qpcpp_5_5_0 Version 5.5.0, 2015-09-28
The main purpose of this release is the extension of the QS software
tracing system to bi-directional communication with embedded Targets.
Specifically, the QS-RX (receive channel for QS) has been added with the
following capabilities:
1. Set global QS filters inside the Target
2. Set local QS filters inside the Target
3. Inject an arbitrary event to the Target (direct post or publish)
4. Execute a user-defined callback function inside the Target with
arguments supplied from QSPY
5. Peek data inside the Target and send to QSPY
6. Poke data (supplied from QSPY) into the Target
7. Execute clock tick inside the Target
8. Request target information (version, all sizes of objects, build
time-stamp)
9. Remotely reset of the Target
This QP/C++ version complements the recent release of Qtools 5.5.0,
where the QSPY host application (https://www.state-machine.com/qspy ) has
been extended with a UDP socket, which is open for communication with
various Front-Ends (GUI-based or headless). An example Front-End written
in Tcl/Tk called "QSpyView" has been developed to demonstrate all the
features. The example application located in the directory
qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl/qspy contains customization of
the "qspyview" script for the DPP application. Please refer to the
documentation of this example
(https://www.state-machine.com/qpcpp/arm-cm_dpp_ek-tm4c123gxl.html ) for
more information.
Also, this release adds a state machine operation for implementing the
shallow history mechanism. The operation is called "childState", because
it computes a child state of a given parent, such that the child belongs
to the same state hierarchy as the current state.
Also, this release fixes problems with QP/C++ examples for Qt. The Qt
project files have been modified to build QP/C++ from sources rather
than using QP as a library.
Finally, this release fixes bugs #110 and #115 reported against QP/C++
on SourceForge.net.
Changes in detail:
1. Modified the QS software tracing component to add new functionality,
such as the QS-RX input channel. Also added new trace records.
2. Added file "qstamp.cpp" (in the qpcpp/include/ folder) to provide
time-stamp of the application build.
3. Added function QMsm::childStateObj() to the QMsm class and
QHsm::childState() to the QHsm class. These functions have been added to
support the shallow-history mechanism.
4. Modified all example projects (qpcpp/examples/ folder) to include the
"qstamp.cpp" file and force its re-compilation for each new build, so
that every build has an up-to-date and unique time stamp.
5. Extended the DPP on TivaC LauchPad example (directory
qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl/) to demonstrate QS-RX (QS
receive channel).
6. Provided example of customizing the "QSpyView" Tcl/Tk script for the
DPP application in the directory
qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl/qspy/
7. Modified all examples (qpcpp/examples/ folder) to call the
QS_ASSERTION() macro to the Q_onAssert() callback function.
8. Modified the startup code (in the qpcpp/3rd_party/ folder) for ARM
Cortex-M to invoke the Q_onAssert() callback from the assert_failure()
exception handler. This is to allow application-level code to define
Q_onAssert() for each specific project.
9. Replaced deprecated registers in TM4C (TivaC) projects
(SYSCTL->RCGCGPIO rather than the deprecated SYSCTL->RCGC2).
10. Fixed bug #110 "QF::init() failed to initialize the last entry in
active_[] array"
11. Fixed bug #115 "QMsm::isInState() implementation missing"
------------------------------------------------------------------------------
@section qpcpp_5_4_2 Version 5.4.2, 2015-06-06
The main focus of this release is to improve the support for "dual targeting" of QP/C++ applications, which is developing of deeply embedded code as much as possible on the desktop OS, such as Windows. Experience shows that "dual targeting" dramatically improves productivity of embedded systems developers, perhaps more than any other technique.
This release makes it possible to use exactly the **same** application code, main function, and the Board Support Package interface (bsp.h) on both deeply embedded target and on Windows. The only differences between these targets can be completely encapsulated in the Board Support Package implementation (bsp.cpp).
The support for "dual targeting" in this QP/C++ release works both for Win32 console and Win32 GUI applications. The Win32-GUI support enables developers to easily emulate the front-panels of the embedded devices, with LCD-screens (graphical and segmented), LEDs, buttons, switches, sliders, etc.
Changes in detail:
1. Modified the QP/C++ ports to Windows (both @ref win32 and @ref win32-qv) so that they support both Win32 console and Win32-GUI applications. The newly introduced pre-processor #WIN32_GUI macro is now required to use the Win32-GUI facilities.
2. Added portable "safe" macros from `<stdio.h>` and `<string.h>` to the QP/C++ ports to Windows. These macros encapsulate the differences between Microsoft Visual C++ and other compilers (such as MinGW).
3. Simplified the structure of the QP/C++ Windows ports by eliminating one level of directories for the compilers used. Both VC++ and MinGW builds can now be run in the same port directory.
4. Modified the QF::stop() function in the QP/C++ port to @ref win32-qv, so that it unblocks the QV event-loop and thus lets the application terminate.
5. Modified all examples for Windows to use the new port structure.
6. Improved all Makefiles (for the MinGW toolset) in all Windows examples, to make them easier to adapt to custom applications, both Win32 console and Win32 GUI.
7. Moved several examples from the `examples/win32/` and examples/win32-qv directories to `examples/arm-cm/` directory with native embedded examples for ARM Cortex-M. This co-location of the Win32 emulation with the embedded code running on the actual board demonstrates better the "dual targeting" development approach.
8. Updated all Windows examples to the latest QP API by compiling the code with the macro #QP_API_VERSION set to 9999 (latest API without backwards compatibility)
9. Improved the PC-Lint support for checking the application-level code located in in `examples/arm-cm/dpp_ek-tm4c123gxl/lint`
------------------------------------------------------------------------------
@section qpcpp_5_4_1 Version 5.4.1, 2015-05-21
This release changes the active object class hierarchy so that ::QMActive is now more fundamental and is the base class for QActive. (Previously ::QMActive was a subclass of ::QActive). The newly added documentation section about @ref classes "QP/C++ Design" shows the current class hierarchy.
@note
Because the types QMActive and QActive are equivalent in QP/C++, this change has minimal impact on the applications, but it is now more correct to use QMActive as the base class for all "opaque" active object pointers.
Also, this release brings several cosmetic improvements:
1. All QM models have been saved with QM 3.3.0, which means that they will not open with QM 3.2.x or earlier QM versions.
2. The `qpcpp/ports/arm-cm/qk/gnu/qk_port.s` ARM Cortex-M port to QK with GNU has been modified to use the CMSIS-compliant symbol __FPU_PRESENT instead of the FPU_VFP_V4_SP_D16 symbol.
3. All Makefiles for the GNU toolset have been cleaned up, whereas any `/` (back-slash) characters in the paths have been repalced with `/` (forward-slash) characters. Also all these Makefiles have been updated to provide the __FPU_PRESENT to C/C++ and assembler when the hardware FPU is used.
4. The file display drver for the EK-LM2S811 board locate at `qpc/3rd_party/ek-lm3s811/display96x16x1.c` has been modified to fix the problem with incorrect hardware delay with the GNU compiler at higher levels of optimization. The in-line assembly for the GNU compiler has been updated such that the delay loop cannot be "optimized away".
5. Several README files have been updated.
------------------------------------------------------------------------------
@section qpcpp_5_4_0 Version 5.4.0, 2015-05-14
This release changes the basic philosophy of distributing the QP frameworks by **combining** the "QP/C++ Baseline Code" with all currently available "QP/C++ Development Kits" (QDK/C++). This is done to eliminate any potential mistakes in downloading and installing separate pieces of code.
Additionally, this release changes the basic philosophy of building your embedded applications with the QP/C++ framework. Starting with this release, all @ref exa "examples" for embedded boards include the QP/C++ framework as **source code** within the projects, instead of statically linking with a QP/C++ library. (**NOTE:** It is still possible to use QP/C++ as a library, but you need to build such libraries yourself, as they are no longer provided in the QP/C++ distribution.)
The move to building QP/C++ from sources ensures the consistent toolset version and compiler options applied to the application code as well as the QP/C++ framework code. (**NOTE:** The QP/C++ examples for "big operating systems", like @ref exa_win32 "Windows" or @ref exa_posix "Linux", still use QP/C++ as a pre-compiled library that is statically linked with the application code.)
This release also changes the active object class hierarchy so that QMActive is now more fundamental and is the base class for QActive. (Previously QMActive was a subclass of ::QActive). The newly added documentation section about @ref classes "QP/C++ Design" shows the current class hierarchy.
@note
Even though the QP/C++ source has been re-packaged in this release, the changes have minimal impact on the existing QP/C++ applications. The biggest difference is that "opaque" pointers to active objects derived from QMActive now need to be also typed QMActive (as opposed to QActive), because QMActive is no longer a subclass of QActive and the automatic upcast no longer applies.
The changes in basic approach to distributing and building the framework have also the following ripple effects:
1. The QP/C++ source code has been simplified and has been re-packaged into a much smaller number of source files. The whole QP/C++ source code now resides in the single <span class="img folder"><a href="dir_b2f33c71d4aa5e7af42a1ca61ff5af1b.html"><strong>source</strong></a></span> folder. Additionally, the source code files have now the **read-only** protection to prevent inadvertent changes to the QP/C++ source code that is part of your projects.
2. It is no longer necessary to define the **QPCPP environment variable** to build the QP/C++ examples. All directories and files referenced by example projects are **relative** to the project folder. This change reflects the fact that most development tools add source files to the project using relative paths (and now the projects contain QP/C++ source code, not just the QP library).
3. The QP/C++ <span class="img folder">@ref ports</span> folder has been reorganized to contain all currently available QP/C++ ports. The ports are organized into three categories: @ref ports_native "native QP/C++ ports" ("bare-metal"), @ref ports_rtos "ports to 3rd-party RTOSes", and @ref ports_os "ports to big operating systems" (Windows and Linux).(**NOTE**: the ports are now documented in the this <strong>QP/C++ Reference Manual</strong>. Each port sub-directory contains a <span class="img qp_link">README</span> link to the corresponding page in the online documentation)
5. The QP/C++ <span class="img folder">@ref exa</span> folder has been reorganized to reduce the repetitions and contains all currently available QP/C++ examples. The folder includes four categories of examples: @ref exa_native "native QP/C++ examples" ("bare-metal"), @ref exa_rtos "examples for 3rd-party RTOSes", @ref exa_os "examples for big operating systems" (Windows and Linux), and @ref exa_mware "examples for 3rd-party Middleware". As mentioned before, all example projects for embedded systems use QP/C++ as source code and not as a library. The <span class="img folder">examples</span> folder has been expanded to contain all currently available QP/C++ examples, many of them are new in this release. (**NOTE**: the currently available examples are now documented in the <strong>QP/C++ Reference Manual</strong>. Each example sub-directory contains a <span class="img qp_link">README</span> link to the corresponding page in the online documentation)
6. A new <span class="img folder">3rd_party</span> folder created to contain the Third-Party code used in the QP/C++ ports and examples, such as MCU register files, low-level startup code, device drivers, etc. The <span class="img folder">3rd_party</span> folder avoids the need to repeat such code in every project. Also, the separation of the Third-Party components helps to clearly indicate code that comes from various sources, and to which Quantum Leaps, LLC expressly makes **no claims of ownership**. The Third-Party software components included in this "3rd_party" folder are licensed under a variety of different licensing terms that are defined by the respective owners of this software and are spelled out in the README.txt or LICENSE.txt files included in the respective sub-folders.
7. This release also comes with the much expanded online <strong>QP/C++ Reference Manual</strong>, which is cross-linked with the ports and examples.
Changes in detail:
1. Renamed the "Vanilla" scheduler to the @ref comp_qv "QV cooperative kernel" for symmetry with the @ref comp_qk "QK preemptive kernel". Renamed QF::onIdle() callback to QV::onIdle().
2. Removed class QFsm (which is now deprecated). Legacy state machines coded in the "QFsm-style" will continue to work, but will use the QHsm implementation internally. There is no longer any efficiency advantage in using the "QFsm-style" state machines.
3. Applied a slight performance improvement to the ARM Cortex-M port to the QK preemptive kernel. The QK port now checks for ISR context by looking at the IPSR register, instead of incrementing and decrementing the `QK_intNest_` up-down counter.
4. Updated @ref exa_arm-cm "ARM Cortex-M examples" and provided new examples for NXP mbed-LPC1768, and STM32 NUCLEO-L053R8, and NUCLEO-L152RE boards. All examples now use the latest CMSIS (V4.3.0). All ARM Cortex-M exampels are provided for the ARM-KEIL, GNU-ARM, and IAR-ARM toolsets.
5. Added the native @ref arm7-9 "port" and @ref exa_arm7-9 "examples" to the @ref arm7-9 "classic ARM7/9" with AT91SAM7S-EK board and the IAR-ARM toolset.
6. Added the native @ref avr "port" and @ref exa_avr "examples" to the AVR (AVRmega) with GNU-AVR and IAR-AVR toolsets. The @ref exa_avr "examples" are provided for the Arduino-UNO board.
7. Added the native @ref msp430 "port" and @ref exa_msp430 "examples" to MSP430 with TI CCS-430 and IAR-430 toolsets. The @ref exa_msp430 "examples" are provided for the MSP430 LauchPad boards (the MSP-EXP430G2 and MSP-EXP430F5529LP for the "classic" MSP430 and "extened" MSP430X, respectively).
8. Added port to @ref cmsis-rtx "CMSIS-RTOS RTX". Examples are available for TI EK-TM4C123GLX, STM32 NUCLEO-L053R8, and NUCLEO-L152RE boards with ARM-KEIL, GNU-ARM, and IAR-ARM toolsets.
9. Updated port to @ref embos "embOS". Examples are available for STM32 STM32F4-Discovery board with IAR-ARM toolset.
10. Updated port to @ref freertos "FreeRTOS" for the latest version 8.2.1. Examples are available for TI EK-TM4C123GLX board with GNU-ARM and IAR-ARM toolsets.
11. Added @ref threadx "port to Thread-X". Example is available for the Thread-X demo with Visual Studio on Windows.
12. Updated port to @ref ucos-ii "uC/OS-II" for the latest version v2.92. Examples are available for TI EK-TM4C123GLX and STM32 NUCLEO-L152RE boards with ARM-KEIL and IAR-ARM toolsets.
13. Updated @ref win32 "port to Win32" (Windows). Modified the port to apply a generous "fudge factor" in over-sizing QP event queues and event pools, to minimize the risk of overflowing queues/pools due to non-deterministic Windows behavior.
14. Updated the @ref qt "QP-Qt Port" (QP/C++ port to the Qt cross-platform GUI framework).
15. Added new @ref win32-qv "port to Win32-QV" (Windows with cooperative "Vanilla" scheduler, previously known as Win32-1T).
16. Updated the @ref lwip_ek-lm3s6965 "lwIP-QP example for EK-LM3S6965 board".
------------------------------------------------------------------------------
@section qpcpp_5_3_1 Version 5.3.1, 2014-09-20
/note QP/C++ 5.3.1 remains backwards-compatible with all QP/C++ ports
and applications
This release fixes the following bugs:
1. QMsm::isInState() returns invalid result (bug #105)
2. QF::gc() doc typo (bug #102)
3. POSIX-port Makefile error (bug #65)
4. Problematic friendship to extern "C" function in qf.h (bug #106)
Also, this release updates the PC-Lint options and removes all the remaining PC-Lint warnings for the latest PC-Lint 9.00k.
Additionally, this release improves the uC/OS-II port in that it is now generic and applicable for any CPU, for which uC/OS-II port exists. Specifically, all references to DOS or x86 have been removed from the QP port and any CPU-specific dependencies have been placed in the separate part of the port.
Finally, this release improves the "QP/C++ Reference Manual" generated by Doxygen and available both inside the QP/C++ baseline distribution (qpcpp.chm file) and online at: https://www.state-machine.com/qpcpp
------------------------------------------------------------------------------
@section qpcpp_5_3_0 Version 5.3.0, 2014-04-14
This release adds the "transition to history" (deep history) feature to
both QHsm and QMsm state machines and their subclasses. This QP/C++
release matches the new QM modeling tool version 3.1.0, which now
supports the "transition to history" connector and the corresponding
code generation for transitions to history.
/note QP/C++ 5.3.0 remains backwards-compatible with QP/C++ applications
developed for QP/C++ 4.x and QP/5.x. However, any QM models created for
the previous QP/C++ versions require re-generating the code with QM 3.1.0.
This release adds new QS (Quantum Spy) instrumentation for tracing
transitions to history as well as entry and exit points in submachines.
All these features require the matching QSPY host application included
in Qtools 5.3.0.
Additionally, the QMsm state machine has been extended to add
implementation of the reusable submachine states and submachines with
entry points and exit points. The reusable submachines in QP/C++ 5.3.0
lay the groundwork for providing reusable submachine states and
submachine diagrams in the next upcoming QM version.
This release also uses the C99 data types uint_fast8_t and uint_fast16_t
instead of the non-standard uint_t.
Finally, this QP/C++ release brings deep changes in the source code
comments and the doxygen documentation generated from the source code.
All comments have now more consistent structure, and every function is
now documented in the implementation file (.cpp file), whereas the
interface (.h files) contain only the brief descriptions of the
functions. This re-structuring of documentation is performed as part of
the validation and verification effort that has begun to provide a
certification package for QP/C++ for safety standards, such as IEC 61508
and ISO 62304 (FDA 510(k)).
Changes in detail:
1. Moved detailed documentation of functions from the header files (.h)
to implementation files (.cpp).
2. Removed the header file "qevt.h" and merged its contents into "qep.h"
3. Added macros: trace records QS_QEP_TRAN_HIST, QS_QEP_TRAN_EP, and
QS_QEP_TRAN_XP to "qs.h"
4. Added macros: Q_TRAN_HIST(), QM_TRAN_HIST(), QM_TRAN_EP(),
QM_TRAN_XP(), and QM_SUPER_SUB() to "qep.h"
5. Added attributes entryAction and initAction to the QMState struct in
"qep.h" (needed for transition to history).
6. Added attribute act to the QMAttr union in "qep.h" (needed for
transitions to entry point in submachine states).
7. Changed the QState return type from action/state handler functions to
uint_fast8_t.
8. Added QM models to the "Transition to History" design pattern example
(directory qpcpp/examples/win32/mingw/history) and also added an example
for "Transition to History" with the QMsm class (see
qpcpp/examples/win32/mingw/history_qm).
9. Changed the m_prio attribute of QActive to uint_fast8_t.
10. Changed the type of prio argument to uint_fast8_t and qlen/stkSize
to uint_fast16_t in the signature of QActiveVtbl.start function pointer
and QActive::start_() implementation.
11. Changed the type of the tickRate argument in QTimeEvt constructor
and ::QF::tickX_(), and ::QF::noTimeEvtsActiveX() to uint_fast8_t.
12. Changed the type of the poolSize argument in ::QF::poolInit() to
uint_fast16_t.
13. Changed arguments evtSize and margin in ::QF::newX_() to
uint_fast16_t.
14. Changed attribute bits in QPSet8 as well as bytes and bits[] in
QPSet64 to uint_fast8_t.
15. Changed the QEQueueCtr event queue counter type to uint_fast8_t.
16. Changed type of arguments qLen and margin in ::QEQueue::init() and
QEQueue/QActive::post() to uint_fast16_t.
17. Changed the return type from QK_schedPrio_() (priority) as well as
the p argument in QK_sched_() and QK_schedExt_() to uint_fast8_t
18. Added function QMsm::isInState() to "qep.h" and its implementation
file qmsm_in.cpp. This function tests whether the QMsm state machine (or
its subclasses like QMActive) "is in" the given state.
19. Updated all make scripts for QP/C++ ports to include the new
qmsm_in.cpp in the QP/C++ library builds.
------------------------------------------------------------------------------
@section qpcpp_5_2_1 Version 5.2.1, 2014-01-06
This release fixes two bugs.
1. In file qmsm_dis.cpp added saving of the action-table into a
temporary variable *before* exiting the current state to the transition
source. Also, changed the signature of the QMsm::msm_tran() helper
function to take the action table as parameter. NOTE: This bug only
affected the Spy configuration and because of this escaped regression
testing. The internal testing process have been updated to test all
build configurations: Debug, Release, and Spy.
2. In file qs_mem.cpp fixed an error in accounting used bytes in the QS
trace buffer.
------------------------------------------------------------------------------
@section qpcpp_5_2_0 Version 5.2.0, 2013-12-28
This release matches the new QM 3.0.0, for which it provides model
examples based on the new QMsm/QMActive classes. This, in turn
demonstrates the new state machine code generation that QM3 was
specifically designed to do.
This release also provides a clearly separated API compatibility layer,
whereas you can configure a level of backwards compatibility by means of
the #QP_API_VERSION macro. This facilitates migrating existing QP
applications to the newer API.
An cyclomatic complexity (McCabe V(G)) analysis of this version has been
performed and the maximum V(G) complexity per function has been reduced
to 15 by breaking up the QHsm::dispatch() function. The code metrics
report, including cyclomatic complexity by function as well as other
standard code metrics (e.g., lines of code), is now included in the
"QP/C++ Reference Manual", see
https://www.state-machine.com/qpcpp/metrics.html
Also, in this release all internal QP data that were previously
uninitialized are now explicitly initialized to zero. In other words,
this release no longer assumes that all uninitialized data (global and
static inside functions) is implicitly initialized to zero before the
control is transferred to main(). This is a requirement of the C
Standard, but some embedded startup code fails to do this.
Finally, this release demonstrates safer stack allocation and safer
exception handlers in all ARM Cortex-M examples. The techniques are
described in the Embedded.com article "Are We Shooting Ourselves in the
Foot with Stack Overflow?".
Changes in detail:
1. for backwards compatibility, in file qp_port.h defined "API
Compatibility Layer", which is controlled by the macro #QP_API_VERSION.
For example, specifying QP_API_VERSION=500 chooses API compatible with
QP version 5.0.0 or newer, but excludes APIs that became deprecated in
the earlier versions. If the macro #QP_API_VERSION is not defined by the
user (typically on the command line for the compiler), the default value
of 0 is assumed. This default means maximum backwards compatibility
(from version 0.0.0). On the other hand, higher values of
#QP_API_VERSION mean less backwards compatibility. For example
QP_API_VERSION=9999 will specify compatibility only with the latest
version of QP. The API Compatibility Layer for QP_API_VERSION < 500
provides macros: postLIFO(), publish(), and tick(). These macros resolve
to POST(), PUBLISH(), and TICK_X() respectively.
2. In file qhsm_dis.cpp, broken up the function QHsm::dispatch() into
two functions QHsm::dispatch() and a private helper QHsm::hsm_tran().
This has reduced the cyclomatic complexity from 25 for the original
function, to 11 and 15 for QHsm::dispatch_() and QHsm::hsm_tran(),
respectively.
3. In file qmsm_dis.cpp, broken up the function QMsm::dispatch() into
two functions QMsm::dispatch() and a private helper QMsm::msm_tran().
This has reduced the cyclomatic complexity from 15 for the original
function, to 9 and 7 for QMsm::dispatch() and QMsm::msm_tran(),
respectively.
4. Changed QM models dpp.qm and game.qm to use QMActive and QMsm base
classes instead of QActive and QHsm, to showcase the new code generation
capabilities of QM3.
5. Added the QMsmTst example application analogous to the QHsmTst to
test the QMsm base class. The QMsmTst comes with the QM model qmsmtst.qm.
6. In file qf_act.cpp added the function QF::bzero(), and in files
qvanilla.cpp and qk.cpp added calls to QF::bzero() to explicitly clear
the uninitialized data. Also added calls to QF::bzero() inside
qf_psini.cpp.
7. Updated all examples for ARM Cortex-M to use safer stack allocation
and safer exception handlers in all ARM Cortex-M examples, as described
in the Embedded.com article "Are We Shooting Ourselves in the Foot with
Stack Overflow?".
------------------------------------------------------------------------------
@section qpcpp_5_1_1 Version 5.1.1, 2013-10-10
This release fixes reversal of logic in the QF::noTimeEvtsActiveX()
function as well as sleep mode transition in the ARM Cortex-M3/M4 ports
to the cooperative Vanilla kernel. Also, the native QP event queue
implementation has been changed to count the extra "front-event"
location into the number of free entries, which fixes the problem of
defer queues of depth 1. Finally, the release restores the support for
linting (with PC-Lint) of the QP/C++ applications for ARM Cortex-M (with
IAR and GNU compilers).
Changes in detail:
1. In file qf_tick.cpp reversed the logic inside QF::noTimeEvtsActiveX()
2. Modified free entry accounting (nFree) in the files: qeq_init.cpp,
qeq_fifo.cpp, qeq_get.cpp, and qeq_lifo.cpp.
3. Modified free entry accounting (nFree) in the files: qa_init.cpp,
qa_fifo.cpp, qa_get_.cpp, and qa_lifo.cpp.
4. Introduced new macro QF_CPU_SLEEP() in the ARM Cortex-M Vanilla ports.
5. Changed Board Support Package files (bsp.cpp) in the ARM Cortex-M
Vanilla examples.
6. Modified the CMSIS-compliant startup code in all ARM Cortex-M Vanilla
examples.
7. Modified the application examples with PC-Lint
(qpcpp/examples/arm-cm/qk/gnu/dpp-qk_ek-lm3s811-lint and
qpcpp/examples/arm-cm/qk/iar/dpp-qk_ek-lm3s811-lint). Updated lint files
for the latest PC-Lint
8. Ported all state pattern examples to Win32/MinGW. That way, all these
examples can be run and modified directly on Windows, without going back
to DOS.
------------------------------------------------------------------------------
@section qpcpp_5_1_0 Version 5.1.0, 2013-10-02
/note QP/C++ 5.1.0 remains <b>backwards-compatible</b> with the existing
QP/C 4.x applications, except the <b>ARM Cortex-M applications</b> need to
adjust the interrupt priorities. Specifically, you need to set the
interrupt priorities equal or lower than <b>QF_AWARE_ISR_CMSIS_PRI</b>
constant provided in the qf_port.h header file.
The main purpose of this milestone QP/C++ release is to enable the QM
modeling tool to generate a new type of state machine code (requires QM
version 3.0.0, which is still in development as of this writing).
This new type of state machine implementation in QP/C++ 5 is based on
the new ::QMsm class, which takes advantage of the QM tool as an advanced
"state machine compiler". QM can perform optimizations that were not
possible with the C pre-processor alone. Specifically, the QM can easily
determine the LCA (Least-Common-Ancestor) state for every transition and
it can generate the complete transition-sequences (sequences of
exit/entry/initial actions) at code-generation time. The resulting code
can be still highly human-readable, but it will no longer be
human-maintainable. The lab tests indicate that the new "housekeeping"
code for executing hierarchical state machines can be about twice as
fast as the previous code based on the ::QHsm class.
The next important change introduced in QP/C++ 5 is making several
important functions in QP/C++ base classes virtual. The new virtual
functions include: QMsm::init() and QMsm::dispatch() and
QActive::start(), QActive::post(), and QActive::postLIFO() perations.
Making theses functions virtual means that all these operations can be
overridden in sub-classes of state machines and active objects. This, in
turn, allows a single application to use a mix of state machine classes
derived from the new QMsm base class with state machines derived from
the QHsm base class, each one using a different state machine
implementation strategy. Additionally, the virtual QActive::post()
operation could be very useful for implementing various Proxy active
objects (e.g., for active object event posting across networks).
Another big feature introduced in QP/C++ 5 are the multiple, independent
system clock tick rates for time events. The number of system tick rates
can be now configured in the QP/C++ ports. For example, a digital watch
can use a "fast" clock tick rate of 100Hz and a "slow" clock tick rate
of only 1Hz. These clock tick rates can be managed independently, so for
example, the fast clock tick rate can be shut down in the absence of
time events assigned to this rate. This feature allows the applications
to implement sophisticated power-saving policies.
As yet another important feature, QP/C++ adds a new "extended" API for
non-asserting event allocation and posting. This feature is intended for
situations, where an application is hammered with external events that
at times arrive too fast for processing, but that can be ignored under
the overload conditions. In those cases firing an assertion inside the
framework is undesirable. The non-asserting API allows a designer to
request a safety-margin when allocating or posting an event. The event
is not allocated/posted if the safety margin cannot be satisfied at the
time of the call. On the other hand, the safety margin allows the
application to still use the regular (asserting) event allocation and
posting, because the event pools and event queues are guaranteed to
maintain a minimal margin for safe operation.
Finally, this QP/C++ release brings significant improvements to the QS
software tracing implementation and also brings important changes the
ARM Cortex-M port. Changes in detail are as follows:
0. Removed the conditional compilation of the QP:: namespace based on
the macro Q_NNAMESPACE. The namespace QP:: is now enabled permanently,
which is a requirement of MISRA-C++:2004 and is also is part of the QM
code generation for QP/C++. Also, the use of explicit QP:: namespace
improves the doxygen documentation (QP/C++ Reference Manual).
1. Added the new QMsm class to qep.h. Changed the inheritance tree by
deriving QHsm and QFsm from the QMsm base class.
2. Added new source files qmsm_ini.cpp and qmsm_dis.cpp to the QEP. These
files implement the QMsm::init() and QMsm::dispatch() functions, respectively.
3. Added the new QMActive class to qf.h. Extended the inheritance
tree to derive QMActive from QActive.
4. Added the multiple system clock tick rates feature in qf.h:
- added new configuration macro #QF_MAX_TICK_RATE, which specifies
the number of clock tick rates. This macro is to be defined in
the QF ports (in the qf_port.h header file). If the macro is
undefined, the default value is 1 (one clock tick rate).
- renamed and re-implemented the QF::tick() function as the
"extended" QF::tickX() function with the argument 'tickRate' for
processing time events allocated to different clock rates. The
application must call QF::tickX(0), QF::tickX(1), ... at the
specified tick rates from ISRs or *tasks*.
- added an "extended" time event constructor, which assigns a time
event to a specific tick rate as well as specific active object.
- renamed and re-implemented the internal function QTimeEvt::arm_()
to a public function QTimeEvt::armX() for arming time events
initialized with the "extended" constructor. The QTimeEvt::armX()
function is the new recommended API for arming time events, both
one-shot and periodic.
- re-implemented QTimeEvt::disarm() and QTimeEvt::rarm().
- renamed QF::noTimeEvtsActive() to the "extended" version
QF::noTimeEvtsActiveX(), which checks time events assigned to the
given tick rate.
5. Added the new non-asserting API to qf.h:
- renamed internal function QF::new_() to QF::newX_(), the latter one
taking the argument 'margin' for allocating events. The function
returns NULL if the event pool has less free events than the
specified margin. The function asserts if the margin is zero and
the event can't be allocated.
- added function QActive::post() to post an event to the given
active object. The function does not post the event if the target
event queue has less free slots than the specified margin. The
function asserts if the margin is zero and the event can't be
posted.
- added "extended" macro Q_NEW_X() for allocating events with a
margin.
- added "extended" macro POST_X() for posting events with a
margin.
6. Modified the QActive::defer() function to return the status of the
defer operation (true==success), instead of asserting when the defer
queue overflows.
7. Modified QS (Quantum Spy) software tracing implementation:
- Optimized the internal QS implementation of all functions that
insert trace data into the trace buffer. The general idea of the
optimization is to extensively use automatic variables instead of
global variables (such as buffer head and tail indexes, the
running checksum, etc.).
- Reduced the number of QS global filters from 256 to 124 (0x7C).
This enables the code to avoid escaping the trace record numbers
- added additional tick rate byte to the trace records QS_QF_TICK
and QS_QFF_TIMEEVT_*.
- An empty QS record and the QS_RESET record are now inserted
automatically into the trace buffer in the function
QS::initBuf(). The empty QS record/QS_RESET pair provides a clean
start of a session and allows the QSPY host application to
re-synch with the data stream.
- added new trace records #QS_QF_ACTIVE_POST_ATTEMPT,
#QS_QF_EQUEUE_POST_ATTEMPT, and #QS_QF_MPOOL_GET_ATTEMPT for the
"extened" non-asserting event allocation and posting.
- added new trace records QS_TEST_RUN and QS_TEST_FAIL for future
support for unit testing.
- added new QS source file qs_dict.cpp with functions QS_*_dict() to
generate various dictionary entries. Changed the macros
QS_*_DICTIONARY() to call these functions. This was done to
significantly reduce the amount of tracing code needed to send the
dictionaries from applications.
- grouped together the various QS variables (such as filters, trace
buffer indexes, etc.) in a single struct, which results in a more
efficient code for various QS operations.
8. Changed the structure of the ARM Cortex-M ports
- renamed the sub-directory for ARM Cortex-M ports and examples
from "arm-cortex" to "arm-cm". This is done to avoid confusion
with other ARM Cortex variants, such as Cortex-A/R, which very
different from Cortex-M.
- Changed the policy of disabling interrupts for Cortex-M3/M4.
For these cores, QP 5.1.0 never completely disables interrupts,
even inside the critical sections. The QP port disables
interrupts selectively using the BASEPRI register. (NOTE: The
BASEPRI register is not implemented in the ARMv6-M architecture
(Cortex-M0/M0+), so Cortex-M0/M0+ need to use the PRIMASK
register to disable interrupts globally).
- removed the CMSIS (Cortex Microcontroller Software Interface
Standard) directories from the Cortex-M examples and moved it to
the common location in the %QPC%/ports/arm-cm/cmsis/ directory.
Upgraded the CMSIS to the latest version 3.20.
- added the ARM Cortex-M ports and examples with Keil/ARM MDK to
the QP Baseline Code.
- upgraded ARM Cortex-M ports with IAR to the latest IAR EWARM 6.60
- upgraded ARM Cortex-M ports with Sourcery CodeBench to the latest
version 2013.05-53.
9. Added the requested simple "Blinky" example for Windows and ARM
Cortex-M (with the GNU, IAR, and Keil toolsets).
- Added "Getting Started with QP/C++" guide based on the Blinky
example.
10. Updated the Doxygen documentation (QP/C++ Reference Manual)
- removed the outdated tutorial section
- updated and added documentation and code samples
- added search box and tree-view panel to the HTML documentation
/note This new policy of disabling interrupts in ARM Cortex-M divides
interrupts into "kernel-unaware" interrupts, which are never disabled,
and "kernel-aware" interrupts, which are disabled in the QP critical
sections. Only "kernel-aware" interrupts are allowed to call QP
services. "Kernel-unaware" interrupts are *NOT* allowed to call any QP
services and they can communicate with QP only by triggering a
"kernel-aware" interrupt (which can post or publish events).
------------------------------------------------------------------------------
@section qpcpp_4_5_04 Version 4.5.04, 2013-02-10
The main purpose of this release is adding support for the ARM Cortex-M4F
processors with the hardware Floating-Point Unit (FPU). The QP/C ports
to Cortex-M4F take full advantage of the "lazy stacking" feature of the
FPU registers, and by doing so offer the most efficient preemptive
multitasking on this processor.
Changes in detail:
1. Added ports and examples for ARM Cortex-M4F with the EK-LM4F120XL
board (Stellaris Launchpad).
2. Added the macro QF_LOG2(), which can be re-implemented in the QP ports,
if the CPU supports special instructions, such as CLZ (count leading zeros
in Cortex-M3/M4). If the macro is not defined in the QP port, the default
implementation uses a lookup table.
3. Updated all ARM Cortex-M ports and examples to the latest IAR EWARM
6.50 and Sourcery CodeBench 2012.09-85.
4. Updated App Notes "QP and ARM Cortex-M with IAR" and "QP and ARM
Cortex-M with GNU".
5. Updated the PC-Lint support files (include/lib-qpcpp.lnt, au-misra2.lnt)
to the latest PC-Lint 9.00j.
6. Updated the Application Note: "QP/C++ MISRA-C:2004 Compliance Matrix".
7. Spell-checked the comments in all QP/C++ source files and removed
several typos.
8. Removed the Qt ports and examples from the QP/C++ Baseline Code and
moved them to the separate QDK/C++-Qt.
/note QP/C++ Version 4.5.04 preserves full compatibility with QM 2.2.03
and all existing QDKs for QP/C++ 4.5.xx.
------------------------------------------------------------------------------
@section qpcpp_4_5_03 Version 4.5.03, 2012-11-29
This release changes the directory structure of QP ports to various
operating systems, such as POSIX (Linux, BSD, etc.), Win32 (Windows),
Qt, etc. The OS ports are moved from the ports/80x86/ directory one
level up to ports/. Also, the OS examples are moved from the
exampels/80x86/ directory one level up to examples/.
------------------------------------------------------------------------------
@section qpcpp_4_5_02 Version 4.5.02, 2012-08-15
The main purpose of this release is providing QM models in most
examples and better, more comprehensive support for (rapid)
prototyping of embedded QP applications on the desktop with the
Win32 API and with Qt. Among others, this release adds a complete
toolkit for creating realistic embedded front panels with pure
Win32-GUI API and free Visual C++ Express and ResEdit. An extensive
Application Note "QP and Win32" is included in this release.
This QP relase adds QM models, created with the new QM 2.2.01,
to most of the examples. The code generated by this new QM version
complies with MISRA-C++:2008 rules. For this compliance, the QM code
generator applies the QP:: namespace prefix to all QP/C++ elements,
such as classes, types, enumerations, etc.
For compliance with MISRA-C++:2008 and compatibility with QM, all
QP/C++ ports in this release use the QP:: namespace. The namespace
generation can be suppressed by providing the new configuration
macro Q_NNAMESPACE, but this is generally *not* recommended, because
it breaks compatiblity with the QM code generation. (In other words,
a QP/C++ port without namespace cannot use the code generated by QM.)
This release simplifies the QP ports to desktop OSs, such as
Windows (Win32), Linux, BSD, Mac OSX (POSIX) and combines 32-bit
and 64-bit ports in one with conditional compilation.
This release also enhances the option for using constructors for
dynamically allocated events. When the configuration macro
Q_EVT_CTOR is defined, the QEvt class provides a non-default
constructor and the Q_NEW() macro becomes variadic and
takes the arguments for the event constructor. This generally
allows creating dynamic events "on-the-fly" without a temporary
pointer to the event. This QP configuration is demonstrated only
in the QP port to Qt, but can be used in any other port.
NOTE: The event constructor feature is NOT backward-compatible
with the existing applications.
At the same time, the QEvt class in this release provides the virtual
destructor only when the new macro Q_EVT_VIRTUAL is defined. This
separation of QEvt constructor from virtual destructor makes the
constructor more attractive, because it saves the virtual pointer
in every event, if no virtual functions are used in QEvt subclasses
(which seems to be generally the case).
This release also adds a new macro QF_MPOOL_EL, which is intended
for allocating properly aligned storage for memory pools and event
pools.
Finally, all QP ports included in this release use only a single QP
library, rather than separate libraries for QEP, QF, QK, and QS.
Changes in detail:
1. Modified QP port to Win32 and used the free Visual C++ Express
2010 with Platform SDK rather than Visual C++ Pro 2008. Renamed
the port directory from vc2008/ to vc/. Provided a completely
revised App Note "QP and Win32".
2. Eliminated QP port to Win32 with one thread (Win32-1T).
3. Consolidated all QP ports to POSIX OSs (Linux, Linux64, Mac_OSX)
into a single port to POSIX and placed it in the directory posix/.
4. Renamed the port directory qt_1t/ to qt/.
5. Added event constructor to qevt.h (controlled by the configuration
macro Q_EVT_CTOR).
6. Added the conditional compilation with the macro Q_EVT_VIRTUAL
around the virtual destructor in the QEvt class.
7. Added macro QF_MPOOL_EL to qmpool.h. Modified all examples to
demonstrate the use of this macro to allocate properly aligned
storage for event pools.
8. Added new typedef 'enum_t' and modified signatures of functions
taking event signals from QSignal to enum_t. This was done to
significantly reduce the number of casts necessary when enumerated
signals were passed to QP functions.
9. Modified all QP ports distributed in the QP/C++ baseline code
to generate only a single QP library, rather than separate
libraries for QEP, QF, QK, and QS. This includes all QP ports
to the desktop (ports/80x86/ directory) and ARM Cortex-M ports
(ports/arm-cortex/ directory).
10. Modified all examples to link only one QP library.
11. Added QM models to most examples and used the automatically
generated code from the models instead of the hand-written code.
12. Modified Qt ports to use the event constructors and
modified examples for Qt to demonstrate this feature.
13. Added .ui files to the Qt examples for generating UIs
graphically with the Qt Designer tool. Revised and updated the
App Note "QP and Qt".
14. Added new macro QS_USR_DICTIONARY() to QS for providing symbolic
names for user-defined trace records
15. Added new macro QS_RESET() to QS for telling the QSPY application
when the target resets. This allows QSPY to reset its internal state.
------------------------------------------------------------------------------
@section qpcpp_4_5_01 Version 4.5.01, 2012-06-14
The main purpose of this minor release is providing improved
MISRA-compliant state machine implementation. Specifically, a new
macro Q_UNHANDLED() has been added for a situation when a guard
condition evaluates to FALSE, but the state model does not prescribe
the explicit [else] case for this guard. In this case, the state
handler can return Q_UNHANDLED(), which will cause the QEP event
processor to propagate the event to the superstate, which is what
UML semantics prescribes.
NOTE: These change to the QEP event processor is completely
backwards-compatible. All state hander functions coded the old
way will continue to handle the guard conditions correctly and
in accordance with the UML specification. The new Q_UNHANDLED()
macro is necessary only for MISRA-compliant state handler coding,
which will be applied in the upcoming release of the QM modeling
and code generation tool.
Changes in detail:
1. Added macro Q_UNHANDLED() and return value Q_RET_UNHANDLED in
qep.h.
2. Modified qhsm_dis.cpp to handle the Q_RET_UNHANDLED return value.
3. Updated the QP/C MISRA-C++:2008 compliance matrix to
include the new MISRA-compliant way of coding guard conditions.
4. Modified qs.h and qs_dummy.h to add new trace record type
QS_QEP_UNHANDLED, which is generated when the state handler returns
Q_RET_UNHANDLED.
5. Modified qs.h and qs_dummy.h to add the User record dictionary
trace record and macro QS_USR_DICTIONARY().
NOTE: This new trace record requires the updated QSPY 4.5.01.
6. Corrected qfsm_dis.cpp, which did not generate QS trace records
for entry and exit from non-hierarchical states.
7. Updated the IAR ARM compiler used in the ARM Cortex-M examples
to the latest version IAR EWARM 6.40.
8. Modified the Qt port not to define the QPApp::onClockTick()
slot function, but instead to allow defining this slot function in
the BSP of the application.
------------------------------------------------------------------------------
@section qpcpp_4_5_00 Version 4.5.00, 2012-05-26
The main pupose of this relase is to improve host-based development of QP
applications, which is critical for Test-Driven Development (TDD). Among
others, this release provides integration between QP and the popular
Qt GUI framework, which allows developers to easily build host-based
simulations of the embedded systems with the realistic user interfaces.
This realase also simplifies implementing transitions to history, which
is a preparation to providing this feature in the QM modeling tool.
Changes in detail:
1. Renamed the event class from QEvent to QEvt to avoid a name conflict
with the Qt framework. Also, for consistency, renamed the file qevent.h
to qevt.h and the macro Q_EVENT_CAST() to Q_EVT_CAST().
NOTE: To minimize impact of this change on the existing QP ports and
applications, the name QEvent is provided as well, but this can be
suppressed by defining the configuration macro Q_NQEVENT in qep_port.h.
2. Changed the design of QF::tick() (file qf_tick.cpp) to better support
calling this function from low-priority tasks (as opposed to interrupts
and highest-priority tasks), which often happens when QP is executed on
the desktop operating systems. In this design only QF::tick() can remove
time events from the active linked list, so no unexpected changes to the
list structure are eliminated
3. Simplified the QTimeEvt class by removing the m_prev link pointer,
as the new design no longer needs a bi-directional list. These changes
impact the files: qte_*.cpp.
4. Added return value to QF::run() to allow transfer of the exit
status to the destop operating systems.
NOTE: This modification haves impact on most QP/C++ ports, because
the QF::run() function must now return a int16_t value.
5. Eliminated the m_running member of the QActive class, which
has been used only in the QP ports to "big" OSes such as Linux
or Windows.
6. Added member m_temp to the QHsm and QFsm base classes to prevent
clobbering the current state (the m_state member) during transitons.
This change allows keeping the current state unchanged during the
entire transition chain, which in turn allows easy and generic access
to the state information to store the state history in the exit
actions from states. Additional bonus of this re-design is the
opportunity of testing for stable state configuration in assertions
added to the qhsm_*.cpp and qfsm_*.cpp files.
7. Added the QHsm::state() and QFsm::state() accessor methods.
8. Modified the "Transition to History" pattern implementation to
use the simplified technique of obtaining the current state in the
exit action from the superstate rather than all the exit actions from
the substates. Modified the "State-Local Storage" (SLS) pattern as
well, because it was using the transition to history constructs.
9. Re-designed the implementation of the QSPY host application, so
that it can be convenienty included as part of the QP library.
This allows direct QS tracing output to the screen for QP applications
running on the desktop.
NOTE: This change is part of the Qtools release 4.5.00.
10. Modified the QP ports to Win32_1t (both the MinGW and VC2008) to
output QS trace data directly to the stdout via the QSPY host-application
interface. Modified the DPP examples for Win32_1T to demonstrate the
direct QS output to the screen.
11. Added QP port to Qt_1t (Qt with one thread) and two example
applications (DPP and PELICAN crossing).
12. Added GNU compiler option -pthread to QP ports for POSIX with
P-threads, including QP ports and examples for Linux and Mac OS X.
------------------------------------------------------------------------------
@section qpcpp_4_4_00 Version 4.4.00, 2012-05-02
The main pupose of this relase is MISRA-C++:2008 compliance, strong-type
checking compliance, update of PC-Lint option files and tests, and
general cleanup.
1. Moved the qp_port.h header file from the port directories to
the qcppp/include/ directory. Also, moved the inclusion of the
QS (Spy) header files (qs_port.h/qs_dummy.h) from qep.h, qf.h,
and qk.h headers to qp_port.h. These structural changes were made
to reduce the number of preprocessor #if nesting levels below 8,
which is the ANSI-C limit. This was done to comply with the MISRA-C
rule 1.1 (all code shall conform to ANSI/ISO C).
NOTE: This modifications have impact on most QP/C++ ports, because
the qp_port.h header file must be removed from the port.
2. Enabled the QP namespace in most QP/C++ ports (except the DOS ports
with the Open Watcom compiler). This modification was needed for
MISRA-C++:2008 compliance with the rule 7-3-1.
NOTE: This modifications have impact on most QP/C++ ports, because
all QP elements require now using the QP:: prefix. However, this prefix
can be dropped when the application applies the "using namespace QP"
directive. (The "using" directive deviates from MISRA-C++:2008 rule
7-3-4, so it is not used in the MISRA-compliant examples. However,
a number of QP ports applies the "using" directive for backwards
compatibilty.)
3. Added the PC-Lint option files std.lnt and lib-qpcpp.lnt to the
qcppp/include/ directory.
4. Cleaned the whole QP/C code from lint comments. All PC-Lint options
have been moved to PC-Lint option files.
5. Modified QP assertion macro Q_DEFINE_THIS_MODULE() to avoid using
the # operator (MISRA rule 16-3-2). This macro now requires the argument
enclosed in doble quotes "".
NOTE: This modification has impact on some QP/C++ ports.
6. Added typedefs for char_t, int_t, float32_t, and float64_t to
event.h header file for compliance with MISRA-C++:2008 rules 3-9-2.
7. Added macros Q_STATE_CAST() and Q_EVENT_CAST() to qep.h to
encapsulate deviation from MISRA-C++:2008 rule 5-2-7.
8. Added macro Q_UINT2PTR_CAST() to encapsulate casting unsigned
integers to pointers, which deviates from MISRA-C++:2008 rules 5-2-7
and 5-2-8. This macro has been added for *application-level* code.
9. Updated ARM Cortex-M examples with the latest CMSIS v3.0, which
complies with more MISRA-C:2004 rules.
10. Added DPP examples for MISRA-C++:2008 compliant applications (for
IAR-ARM and GNU-ARM).
11. Updated ARM-Cortex-M3 port with GNU to the latest Sourcery
CodeBench 2011.09-60.
12. Added QP/C++ port to Win32-1t and examples (Windows with 1 thread).
This port is useful for testing embedded QP/C++ applications on windows.
13. Added documentation to QP/C++ distribution in the directory
qpcpp/doc/, with the following Application notes:
"MISRA-C++:2008 Compliance Matrix", "Quantum Leaps Coding Standard",
"QP and ARM Cortex-M, and QP and Windows",
------------------------------------------------------------------------------
@section qpcpp_4_3_00 Version 4.3.00, 2011-11-03
1. This release changes the names of critical section macros and
introduces macros for unconditional interrupt disabling/enabling.
This is done to simplify and speed up the built-in Vanilla and QK
kernels, which no longer are dependent on the interrupt locking
policy.
NOTE: The change in handling the critical section in the Vanilla and
QK kernels can break QP ports, which use the "save and restore
interrupt lock" policy, because all such ports must also define
unconditional interrupt disabling and enabling.
2. This release changes the partitioning of the QK scheduler.
Specifically, the QK scheduler is now divided between two functions
QK_schedPrio_() and QK_sched_(), to calculate the highest-priority
task ready to run and to perform scheduling, respectively. The function
QK_schedPrio_() is useful to determine if scheduling is even necessary.
3. Updated all QP ports to comply with the new critical section
names and policies.
4. Modified the ARM Cortex-M port qk_port.s to take advantage of the
new structure of the QK scheduler.
5. Upgraded the examples for ARM Cortex with IAR EWARM to the
latest IAR EWARM version 6.30.
6. Upgraded the examples for ARM Cortex with GNU (CodeSourcery) to the
latest Sourcery CodeBench 2011.07-60.
------------------------------------------------------------------------------
@section qpcpp_4_2_04 Version 4.2.04, 2011-09-24
The main pupose of this relase is to provide a bug fix for the QK port
to ARM Cortex processors. The bug fix addresses a very rare and
undocumented behavior of late-arrival of an interrupt while entering
the PendSV exception. In this case the PENDSVSET bit in the NVIC-ICSR
register is *not* cleared when finally PendSV is entered, so the
PendSV exception is entered in a *different* state when it is entered
via the late-arrival mechanism versus the normal activation of the
exception through tail-chaining. The consequence of this undocumented
and inconsistent hardware behavior, PendSV could be re-entered again
before the SVCall exception cleans up the stack. The bug fix is
implemented in the qk_port.s file and consists of clearing the
PENDSVSET bit programmatically inside PendSV_Handler.
------------------------------------------------------------------------------
@section qpcpp_4_2_02 Version 4.2.02, 2011-09-08
1. The main pupose of this relase is to repackage the default QP/C++
distribution to contain the single root directory qpcpp/ in the
archive. That way, unziping the archive will produce only one
directory (qpcpp/), which can be then changed by the user.
2. This release also changes the ARM Cortex QP ports with GNU. The
suffix "_cs" has been added to all QP libraries generated by the
Code Sourcery toolset (now Mentor Sourcery CodeBench). This is to
distinguish libraries generated by different GNU-toolchains (such
as CodeRed, Attolic, DevKit ARM, etc.)
------------------------------------------------------------------------------
@section qpcpp_4_2_01 Version 4.2.01, 2011-08-13
1. Modified file qassert.h to add assertion macros #Q_ASSERT_ID,
#Q_REQUIRE_ID, #Q_ENSURE_ID, #Q_INVARIANT_ID, and #Q_ERROR_ID,
which are better suited for unit testig, because they avoid the
volatility of line numbers for indentifying assertions.
2. Added QP port and examples for Mac OS X on 80x86.
------------------------------------------------------------------------------
@section qpcpp_4_2_00 Version 4.2.00, 2011-07-15
The goal of this milestone release is to extend the number of event
pools (theoretically up to 255 pools) instead of just three event
pools available up til now. This release adds an option to use
constructors/destructors and virtual functions in subclasses of QEvent.
Also, this release adds an option to wrap the whole QP framework in
the QP namespace to avoid name conflicts with other libraries and to
improve QP compliance with MISRA C++ 2008. This release adds also
several improvements to the QS/QSPY software tracing, such as adding
sender information to event posting so that sequence diagrams could be
easily and automatically reconstructed from tracing data. Also, the
tracing now supports 64-bit targets, such as embedded Linux 64-bit.
Finally, this milestone release migrates the examples to use the
environment variable QPCPP instead of relying on the relative path to
the QP/C++ framework. This allows easier adaptation of the examples for
real projects, which don't really belong to the examples directory.
The changes in detail are:
1. Changed the QEvent base class (file qevent.h). The private member
'dynamic_' has been replaced by two members 'poolId_' and 'refCtr_'.
2. Added conditionally-compiled constructor and virtual destructor
to the QEvent base class (file qevent.h). Also added the inclusion
of the <new> header file for placement new if the constructor/
destructor option is configured.
3. Added new version of the macro #Q_NEW() with variable number of
paramters, which invokes the event constructor (via placement new)
with any parameters requried by the event constructor (file qf.h).
4. Added explicit call to the event destructor in QF::gc() (file
qf_gc.cpp).
5. Added configuration macro QF_MAX_EPOOL (file qf.h) to define the
maximum number of event pools in the QP application (default to 3).
The maximum theoretical number of this macro is 255.
6. Made algorithmic changes in the QF source code related to the change
of storing the event pool-IDs and reference counters inside QEvent.
7. Changed the default signal size (macro Q_SIGNAL_SIZE in the file
qevent.h) from 1 to 2 bytes.
8. Changed the signature of QActive::recall() to return uint8_t instead
of QEvent*, which this could encourage incorrect usage (processing of
the event at the point of recalling it). Now, the function only returns
1 (TRUE) if the event was recalled and 0 (FALSE) if the event was not
recalled.
9. Added the QTimeEvt() constructor and new source file qte_ctr.cpp.
The function returns the counter of a time event, which allows using
a time event for measuring the time.
10. Added new QF macros #QF_TICK, #QF_PUBLISH, and #QACTIVE_POST in
file qf.h to provide sender of the events for software tracing.
11. Added new QS macros (files qs.h and qs_dummy.h) for handling 64-bit
integers.
12. Added the functions QS::u64() and QS::u64_() and new source file
qs_u64.cpp.
13. Added the QS macro #QS_U32_HEX_T for hexadecimal formatting of
integer numbers in the user-defined trace records.
14. Added the new port linux64 for 64-bit Linux. Also added the
corresponding examples for 64-bit Linux.
15. Adapted the QSPY host application for 64-bit pointer sizes and
the changed layout of trace records that contain event information
(such as PoolID and RefCtr). Also added the backwards-compatibility
option (-v) for switching the tool to the previous data format.
16. Removed the tools directory from the QPC distribution and moved
the QSPY host application to the QTOOLS distribution, which now also
contains the GNU make tools for Windows.
17. Modified the make files and project files to use the environment
variable QPCPP instead of relying on the relative path to the QP/C++
framework.
18. Upgraded the examples for ARM Cortex with IAR EWARM to the
latest IAR EWARM 6.20.
------------------------------------------------------------------------------
@section qpcpp_4_1_07 Version 4.1.07, 2011-02-28
The goal of this release is to improve the ease of experimenting with
QP/C++ on the desktop. This release adds support for Windows (Win32) to
the baseline code. Two most popular compilers for Windows are supported:
Microsoft Visual Studio and MinGW (GNU). The support for Linux has been
improved by including pre-built QP/C++ libraries and improving makefiles
for Eclipse compatibility.
The changes in detail are:
1. Added Win32 port with the Visual C++ 2008 (ports/80x86/win32/vc2008).
This directory contains the Visual Studio solution all_qp.sln for
building all QP/C++ libraries from the IDE. Three build configurations
(Debug, Release, and Spy) are supported.
2. Added Win32 port with the MinGW comiler (ports/80x86/win32/mingw).
This directory contains the Makefile for building all QP/C++ libraries.
Three build configurations (dbg, rel, and spy) are supported.
NOTE: the Makefile assumes that the MinGW/bin directory is added
to the PATH.
3. Added Win32 examples for Visual C++ 2008 (examples/80x86/win32/
vc2008/dpp and examples/80x86/win32/vc2008/qhsmtst). Visual Studio
soultions are provides for all build configurations.
4. Added Win32 examples for MinGW (examples/80x86/win32/mingw/dpp
and examples/80x86/win32/mingw/qhsmtst). Eclipse-compatible makefiles
are provided for all build configurations. NOTE: the Makefiles assume
that the MinGW/bin directory is added to the PATH.
5. Removed memory alignment correction in the file qmp_init.c. This
correction required casting of pointers to integers and was problematic
on 64-bit targets (such as 64-bit Linux).
6. Upgraded the examples for ARM Cortex with CodeSourcery to the
latest Sourcery G++ 2011.02-2.
------------------------------------------------------------------------------
@section qpcpp_4_1_06 Version 4.1.06, 2011-01-07
1. Fixed a bug in the software tracing instrumentation in the function
QF::remove_(), file qf_act.cpp.
2. Made cosmetic improvements to the example QM models of the
"Fly 'n' Shoot" game.
3. Made improvements in make.bat files for building the examples for
DOS/Open Watcom to run better in DosBox on Linux.
4. Upgraded the examples for ARM Cortex with IAR to the latest
IAR EWARM version 6.10.
5. Upgraded the examples for ARM Cortex with CodeSourcery to the
latest Sourcery G++ 2010.09-66.
------------------------------------------------------------------------------
@section qpcpp_4_1_05 Version 4.1.05, 2010-11-01
This release is adds examples for the QM (QP Modeler) graphical modeling
and code generation tool. The examples are based on the "Fly 'n' Shoot"
game described in the QP/C++ Tutorial and in Chapter 1 of the PSiCC2 book.
Specifically, the directory <qpcpp>/examples/80x86/dos/watcom/l/game-qm/
contains the "Fly 'n' Shoot" game model file "game.qm". This model, when
opened in the QM tool contains all state machine diagrams and generates
code into the subdirectory qm_code/. This code can then be built and
executed on any 80x86 machine (newer versions of Windows or Linux require
the DOSbox application, see http://www.dosbox.com).
The directory <qpcpp>/examples/arm-cortex/vanilla/iar/game-ev-lm3s811-qm/
contains the version of the game for the EV-LM3S811 ARM Cortex-M3 board.
This directory contains the model file "game.qm", which is actually
identical as the model in the DOS version. The LM3S811 version needs to
be compiled with the IAR compiler and executed on the EV-LM3S811 board.
Additionally, the QP/C++ baseline code has been slighlty modified for
better conformance to the MISRA C++ 2008 rules and the latest PC-Lint 9.x.
------------------------------------------------------------------------------
@section qpcpp_4_1_04 Version 4.1.04, 2010-05-16
This release is adds compatibility of all examples for DOS with the DOSBox
emulator (http://www.dosbox.com/) that recreates a MS-DOS compatible
environment on all versions of Windows, including 64-bit Windows that don't
run 16-bit DOS applications anymore.
Also, this release includes QP ports and examples for EV-LM3S811 board with
the GNU-based Code Sourcery G++ toolset. Support for Sourcery G++ provides
a very cost-effective option for developing QP applications for ARM Cortex
MCUs.
Finally, this release improves the Cortex Microcontroller Software Interface
Standard (CMSIS) for the whole family of the Stellaris LM3Sxxx MCUs. The
improvement extends the CMSIS from Sandstorm to Fury, DustDevil, and Tempest
Stellaris families.
------------------------------------------------------------------------------
@section qpcpp_4_1_03 Version 4.1.03, 2010-01-21
This release is concerned with the ARM Cortex ports and examples.
Specifically, this release contains the following improvements:
1. Unified source code for ARM Cortex-M3 and the new ARM Cortex-M0 cores,
including the code for the preemptive QK kernel.
2. Compliance with the Cortex Microcontroller Software Interface Standard
(CMSIS) in all ARM Cortex examples.
3. Backward-compatible support for the new LM3S811 EVAL Rev C board with
different OLED display than previous revisions. (NOTE: The OSRAM 96x16x1 OLED
used in REV A-B boards has been replaced RITEK 96x16x1 OLED used in Rev C.)
In the process of making the examples CMSIS-compliant, the dependency on the
Luminary Micro driver library (driverlib.a) has been completely removed.
Additionally, the screen saver of the "Fly 'n' Shoot" game has been improved
to periodically switch off the power of the OLED display, which better
protects the display from burn-in. The affected file is tunnel.cpp.
Finally, this release introduces the QP_VERSION macro, which identifies the QP
version. Otherwise, this maintenance release does not change the QP/C API in
any way, so the release has NO IMPACT on the QP/C applications except for the
ARM Cortex ports and applications.
------------------------------------------------------------------------------
@section qpcpp_4_1_02 Version 4.1.02, 2010-01-14
The purpose of this minor maintenance release is the change in the directory
structure for the ARM Cortex ports and examples. As new ARM Cortex cores are
becoming available, the old port name "cortex-m3" could be misleading, because
it actually applies to wider range of ARM Cortex cores. Consequently, all ARM
Cortex ports and examples are hosted in the directory tree called
"arm-cortex".
This maintenance release does not change the QP/C++ API in any way, so the
release has NO IMPACT on the QP/C++ applications except for the ARM Cortex
ports.
------------------------------------------------------------------------------
@section qpcpp_4_1_01 Version 4.1.01, 2009-11-05
The main purpose of this release is to replace the Turbo C++ 1.01 toolset
with the Open Watcom C/C++ toolset, because Turbo C++ 1.01 is no longer
available for a free download. In contrast, Open Watcom is distributed under
an OSI-certified open source license, which permits free commercial and
non-commercial use. Open Watcom can be downloaded from www.openwatcom.org.
All 80x86/DOS and 80x86/qk ports and examples for Turbo C++ 1.01 have been
replaced with ports and examples for Open Watcom. The make.bat scripts are
provided to build the QP/C++ libraries and examples.
In the process of converting the examples to Open Watcom two new examples
have been added to the standard QP/C++ distribution. The Calc2 example
located in &lt;qpcpp&gt;/examples/80x86/dos/watcom/l/calc2 shows how to derive
state machine classes with QP 4.x. The SLS example located in
&lt;qpcpp&gt;/examples/80x86/dos/watcom/l/sls shows the implemenation of the
new State-Local Storage state design pattern.
------------------------------------------------------------------------------
@section qpcpp_4_1_00 Version 4.1.00, 2009-10-09
The release provides brings a number of improvements to QP/C++ and updates
the QP/C++ ARM Cortex-M3 examples for the EK-LM3S811 board to the latest
IAR EWARM 5.40. Due to the acquisition of Liminary Micro by Texas Instruments
the directory IAR structure for the examples and drivers has changed and the
QP examples had to be changed accordingly.
This maintenance release does not change the QP/C++ API in any way, so the
release has NO IMPACT on the QP/C++ applications.
The main changes in QP v4.1.00 with respect to earlier versions are as
follows:
- in qs.h added a new trace record QS_QEP_DISPATCH logged when an event is
dispatched to a state machine. This timestamped trace record marks the
beginning of an RTC step. The end of the RTC step is marked by the existing
timestamped trace records QS_QEP_INTERN_TRAN, QS_QEP_INIT_TRAN, and
QS_QEP_IGNORED, depending on how the event is handled. The new QS_QEP_DISPATCH
record facilitates measurement of the RTC step lengths.
- in qhsm_dis.cpp added generation of the QS_QEP_DISPATCH trace record.
- in the tools/qspy/ sub-directory added output of the new trace record to the
Q-SPY host application.
- in the tools/qspy/matlab/ sub-directory added processing of the new trace
record to the qspy.m script.
- in qpset.h changed the implementation of the Priority Set. In particular,
the QPSet64 now derives from QPSet8, which enables a common way of testing for
non-empty set (e.g., useful in assembly). Also, the findMax() functions in
QPSet8 and QPSet64 now can work with an empty set, in which case they return
0.
- in qk_sched.cpp simplified the QK_schedule_() function to skip the testing
of the ready-set for non-empty condition. Such test is no longer necessary.
The test can still be performed outside the QK_schedule_() function (e.g., in
assembly) to avoid calling the scheduler if the ready set is empty.
- in qk_ext.cpp simplified the QK_scheduleExt_() function in the same way as
QK_schedule_().
- modified make.bat files for building QP libraries in the ports/ directory
to use the externally defined environment variables for the location of the
toolchains. The defaults are applied only when the environment variable is
not defined. This enables greater flexibility in installing the development
tools in different directories than those chosen by Quantum Leaps.
- modified the ARM Cortex-M3 examples for the new IAR EWARM 5.40.
- modified slighlty the Calculator example to allow extensibility.
- in the ARM Cortex-M3 port file qk_port.s added explicit testing of the
QK_readySet_ set for empty condition. This test allows avoiding calling the QK
scheduler and two contex-switches if the ready-set is empty.
- in the game example moved setting up the QS filters from main.cpp to bsp.cpp.
------------------------------------------------------------------------------
@section qpcpp_4_0_04 Version 4.0.04 (Product) Apr 09, 2009
The maintenance release introduces the virtual destructor in the QHsm and
QFsm base classes. This is done to allow proper handling of virtual functions
in the subclasses of QHsm or QFsm. Please note, however, that the change
increases the size of every state machine object by the additional (hidden)
virtual pointer (vptr).
Also, this release provides a fix for the compile-time assertions, which did
not work correctly for the GNU compiler family.
Finally, the ARM Cortex-M3 examples have been recompiled with the newer
IAR EWARM v5.30.
This maintenance release does not change the QP/C++ API in any way, so the
release has NO IMPACT on the QP/C++ applications.
The main changes in QP v4.0.04 with respect to earlier version are as follows:
- in qep.h file, virtual destructors have been added to QFsm and QHsm.
- in qassert.h file the #Q_ASSERT_COMPILE macro has been modified to render
a negative array dimension when the asserted condition is not TRUE.
------------------------------------------------------------------------------
@section qpcpp_4_0_03 Version 4.0.03, 2008-12-27
The main purpose of this release is to fix a bug in the QK preemptive kernel,
which occurs only when the advanced QK features are used. Specifically, the QK
priority-ceiling mutex could interfere with QK thread-local storage (TLS) or
QK extended context switch. When the QK mutex is in use, the TLS or the
extended context for this task could get saved to an incorrect priority level.
The release 4.0.03 fixes the bug by strictly preserving the semantics of
QK_currPrio_ variable. The mutex locking now uses a different variable
QK_ceilingPrio_, which represents the ceiling-priority locked by the mutex.
The QK scheduler and extended scheduler now perform an additional check to
make sure that only tasks with priorities above the ceiling can run. To avoid
that additional overhead, the user can define the macro QK_NO_MUTEX, which
eliminates the QK mutex API and eliminates the additional tests in the QK
schedulers.
This maintenance release does not change the QP/C++ API in any way, so the
release has NO IMPACT on the QP/C++ applications.
The main changes in QP v4.0.03 with respect to earlier version are as follows:
- in qk.h file, made the QK mutex API only visible when the macro QK_NO_MUTEX
is *not* defined.
- in qk_pkg.h file, added the QK_ceilingPrio_ external declaration when the
macro QK_NO_MUTEX is not defined.
- in qk_mutex.cpp file, changed priority-ceiling mutex implementation to use
the QK_ceilingPrio_ instead of QK_currPrio_. Also, added compiler error when
the macro QK_NO_MUTEX is definedthis and this file is included in the build.
- in qk_sched.cpp file added testing priority against the QK_ceilingPrio_,
when the macro QK_NO_MUTEX is not defined.
- in qk_ext.cpp file added testing priority against the QK_ceilingPrio_,
when the macro QK_NO_MUTEX is not defined.
------------------------------------------------------------------------------
@section qpcpp_4_0_02 Version 4.0.02, 2008-11-15
This maintenance release does not change the QP/C API in any way, so the
release has NO IMPACT on the QP/C++ applications.
The main changes in QP v4.0.02 with respect to earlier version are as follows:
- in qep.h file, added comments to macros Q_TRAN() and Q_SUPER() to suppress
the PC-lint warining about using the comma-operator (MISRA rule 42).
- in qhsm_in.cpp file, fixed a bug in the QHsm::isIn() function. - fixed a bug
in tunnel.cpp file ("Fly 'n' Shoot" game). The constant event HIT_WALL was not
declared static.
------------------------------------------------------------------------------
@section qpcpp_4_0_01 Version 4.0.01, 2008-06-09
This maintenace release is made to allow using QS software tracing with
the GNU compiler for AVR (WinAVR). Specifically, the output of the strings
residing in ROM has been fixed.
This maintenance release does not change the QP/C++ API in any way, so the
release has NO IMPACT on the QP/C++ applications.
The main changes in QP v4.0.01 with respect to earlier version are as follows:
- in qs_.cpp file, updated the function QS::str_ROM_().
- in qs_str.cpp file, updated the function QS::str_ROM().
- in qvanilla.cpp file, funciton QF::run(), declared the temporary variables
as static to save stack space, because QF::run() never returns and is not
reentrant.
------------------------------------------------------------------------------
@section qpcpp_4_0_00 Version 4.0.00, 2008-04-07
This milestone release is made for the book /ref PSiCC2. The book describes in
great detail this new release. The older "QP Programmer's Manual" is now
phased out and is replaced with this hyper-linked
/ref main_page "QP/C++ Reference Manual", which provides very detailed, easily
searchable reference to the software. The book /ref PSiCC2 provies in-depth
discussion of the relevant concepts as well as the design study of QP v4.0.
The main changes in QP v4.0 with respect to earlier versions are as follows:
- the coding techniques for hierarchical state machines (HSMs) and the simpler
finite state machines (FSMs) have changed. While the changes are quite simple,
the backward compatibility with QEP 3.x has been broken, meaning that some
manual changes to the state machines implemented with earlier versions are
necessary. Please refer to the "QP/C++ Tutorial" Section /ref coding_hsm for
more information about coding state machines with QEP 4.x.
- The main change is the signature of a state-handler function, which now
returns simply a byte. This return type (typedef'ed as QState) is the
status of the event-handling that the state handler conveys to the QEP
event processor.
- The macro #Q_TRAN() must now always follow the return statement.
- The new macro #Q_SUPER() designates the superstate of the given state.
Again, this macro must follow the return statement.
- Then two new macros #Q_HANDLED() and #Q_IGNORED() have been added to
return the status of event handled and event ignored, respectively.
- all callback functions are now consistently called /c XXX_onYYY() or
/c XXX::onYYY():
- Q_assert_handler() is now Q_onAssert()
- QF::start() is now QF::onStartup()
- QF::cleanup() is now QF::onCleanup()
- the new header file qevent.h has been broken off the qep.h header file.
qevent.h contains the QEvent class and other basic facilities used in the
whole QP. This new file allows easier replacement of the entire QEP event
processor with custom event processors, if you wish do so.
- the macro QEP_SIGNAL_SIZE is renamed to #Q_SIGNAL_SIZE.
- the data type QSTATE is now deprecated. Please use QP::QState.
- the #QF_FSM_ACTIVE macro is now deprecated. Instead, please define
#QF_ACTIVE_SUPER_ to the base class of QP::QActive and #QF_ACTIVE_STATE_ to the
data type represented state. By default, these macros are defined to use the
QP::QHsm base class from the QEP hierarchical event processor, but you can
replace the event processor, if you wish.
- the internal macro QACTIVE_OSOBJECT_WAIT_() is now #QACTIVE_EQUEUE_WAIT_().
- the internal macro QACTIVE_OSOBJECT_SIGNAL_() is now
#QACTIVE_EQUEUE_SIGNAL_().
- the internal macro QACTIVE_OSOBJECT_ONIDLE_() is now
#QACTIVE_EQUEUE_ONEMPTY_().
- the data mebers QActive::m_osObject and QActive::m_thread are now present
only if the macros #QF_OS_OBJECT_TYPE and #QF_THREAD_TYPE are defined.
- the QP::QPSet class has been renamed to QP::QPSet64.
- the QPSet::hasElements() has been renamed QPSet64::notEmpty()
- the QS software tracing is now better integrated with all QP components. You
no longer need to explicitly include qs_port.h, because it is automatically
included when you define the macro #Q_SPY. Also the file qs_dummy.h is
included automatically when the macro #Q_SPY is <STRONG>not</STRONG> defined.
- the new header file qvanilla.h now replaces the file qsched.h.
- the macros QF_SCHED_LOCK() and QF_SCHED_UNLOCK() are now obsolete.
- the native QF event queues (both the active object event queues and the
"raw" thread-safe queues) are slightly more efficient by counting down the
/c head and /c tail pointers rather than up. This leads to wrap-around at
zero, which is easier (faster) to test than any other wrap-around point.
Also the native QF event queues maintain the minimum of the free events in the
queue rather the maximum of used events.
- the data member of QEQueue::nTot class is removed.
- the QF::publish() function has been re-written so that QF no does <STRONG>not
need to lock the scheduler</STRONG>. The QF::publish() function posts events
to active objects with scheduler unlocked starting from the highest-priority
active objects. However, the event is protected from inadvertent recycling by
incrementing its reference counter before the publish operation. After the
event is posted to all subscribers, the garbage collector QF::gc() is called
to decrement the reference counter and recycle the event, if necessary.
- the qf_run.cpp file is obsolete. The QF::run() function for the cooperative
"vanilla" kernel is now implemented in the file qvanilla.cpp.
- the QF::tick() function has been rewritten to allow calling QF::tick() from
the task context as well as from the interrupt context. The nested critical
section around QF::tick() is no longer needed when it is called from the task
level. Among others, this re-design <STRONG>eliminates the need for the
recursive</STRONG> mutex in the POSIX QF port.
- the QMPool::init() function has been re-designed to optimally align the
memory buffer in a portable and platform-independent way. This should bring
some performance improvements on some CPUs (e.g., 80x86).
- the extended QK scheduler has been re-designed to save stack space. The
extended context (e.g., coprocessor registers) are no longer saved on the
precious stack, but rather in the active object.
- a bug has been fixed in handling of Thread-Local Storage (TLS) in the QK
scheduler and extended scheduler.
- the -q (quiet) flag has been added to the QSPY host application.
- the support for two new compilers for Windows has been added for the QSPY
host application. The application can now be build with the MinGW GNU compiler
for Windows as well as the Microsoft Visual C++ 2005.
- the QP port to Linux has been improved by eliminating the need for recursive
P-Thread mutex.
- the QP port to MicroC/OS-II has been upgraded to the latest version 2.86.
- all examples in the standard QP distribution have been cleaned up and
updated to the latest QP API changes.
- all examples that use QF now contain the QS software tracing support.
*/
} // namespace QP
Reference in New Issue View Git Blame Copy Permalink