Version 5.8.2, 2017-02-08
This release adds examples for the ARM Cortex-M7 CPU. Specifically, the release contains the standard Dining Philosophers Problem (DPP) examples for the 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.
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
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 "Setting ARM Cortex-M Interrupt Priorities in QP 5.x".
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.
- 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:
- 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.
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 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).
- 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 bug#147. 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:
- bug#147 "QXK: PendSV_error is triggered on special conditions"
- bug#144 "Obsolete Win32 API in qwin_gui.c"
- bug#124 "Windows port now cause memory leakage"
Version 5.7.4, 2016-11-04
This release fixes the following bugs:
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:
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).
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.
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 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:
- Modified qep_msm.cpp to correct the support for sub-machines and sub-machine states
- Added new port to ARM Cortex-R in the directory ports/arm-cr
- Added examples for ARM Cortex-R in the directory examples/arm-cr
- Modified the ARM Cortex-M QK ports (ARM-KEIL, GNU, IAR, and TI)
- Modified the ARM Cortex-M QXK ports (ARM-KEIL, GNU, IAR, and TI)
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 http://www.state-machine.com/products/index.html#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 (emWin Embedded GUI).
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:
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
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.
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:
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:
- 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.
- Modified QV kernel to provide (dummy) implementation of selective scheduler locking.
- Modified QK kernel to implement selective scheduler locking via modified priority-ceiling mutex QMutex.
- Modified QXK kernel to implement selective scheduler locking via modified priority-ceiling mutex QXMutex.
- 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.
- 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.
- 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.
- 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).
- Modified the Win32 port to provide (global) scheduler locking, which is implemented by Win32 critical section.
- 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().
- 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.)
- Modified QK and QXK examples in qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl to demonstrate the usage of the new priority-ceiling mutexes.
- Fixed the 3rd-party file startup_stm32l32l1xx.c to include exceptions for Cortex-M3 (MemManage_Handler, BusFault_Handler, and UsageFault_Handler).
- Updated the 3rd-party files for the EK-TM4C123GXL board (TivaC LaunchPad).
- 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.
- Implemented Feature Request #110 as well as the duplicate Request #62 by adding function QMActive::flushDeferred()
Version 5.6.1, 2016-01-01
This release is the first official (production) release of the new blocking 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.
- The QXK documentation is available in the QP/C Reference Manual at 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:
- Added new header files for QXK: qxk.h, and qxthread.h.
- Added new source files for QXK: qxk.cpp, qxk_mutex.cpp, qxk_pkg.h, qxk_sema.cpp, qxk_xthr.cpp.
- Added QXK ports to ARM Cortex-M for ARM-KEIL, GNU-ARM, IAR, and TI-ARM toolsets (see qpcpp/ports/arm-cm/qxk)
- 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.
- Removed Q_ROM, Q_ROM_BYTE, and Q_ROM_VAR from the QP/C++ code.
- Added Q_ROM, Q_ROM_BYTE to the compatibility-layer in qpcpp.h.
- 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.
- Removed AVR ports and examples.
- Re-designed the QK priority-mutex in files qk.h and qk_mutex.cpp.
- Provided QK mutex examples in qpcpp/examples/arm-cm/dpp_ek-tm4c123gxl/qk and qpcpp/examples/arm-cm/dpp_nucleo-l053r8/qk.
- Updated Makefiles for GNU-ARM to use the __ARM_ARCH macro for defining the ARM architecture.
- Updated CMSIS from 4.2 to 4.3 in qpcpp/3rd-party/CMSIS
- 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.
- Corrected example projects for the ARM Cortex-M with TI/CCS toolset
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.
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:
- Set global QS filters inside the Target
- Set local QS filters inside the Target
- Inject an arbitrary event to the Target (direct post or publish)
- Execute a user-defined callback function inside the Target with arguments supplied from QSPY
- Peek data inside the Target and send to QSPY
- Poke data (supplied from QSPY) into the Target
- Execute clock tick inside the Target
- Request target information (version, all sizes of objects, build time-stamp)
- Remotely reset of the Target
This QP/C++ version complements the recent release of Qtools 5.5.0, where the QSPY host application (http://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-cm-tm4c123gxl contains customization of the "qspyview" script for the DPP application. Please refer to the documentation of this example (http://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:
- Modified the QS software tracing component to add new functionality, such as the QS-RX input channel. Also added new trace records.
- Added file "qstamp.cpp" (in the qpcpp folder) to provide time-stamp of the application build.
- 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.
- Modified all example projects (qpcpp\ 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.
- Extended the DPP on TivaC LauchPad example (directory qpcpp-cm-tm4c123gxl) to demonstrate QS-RX (QS receive channel).
- Provided example of customizing the "QSpyView" Tcl/Tk script for the DPP application in the directory qpcpp-cm-tm4c123gxl\
- Modified all examples (qpcpp\ folder) to call the QS_ASSERTION() macro to the Q_onAssert() callback function.
- 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.
- Replaced deprecated registers in TM4C (TivaC) projects (SYSCTL->RCGCGPIO rather than the deprecated SYSCTL->RCGC2).
- Fixed bug #110 "QF::init() failed to initialize the last entry in
- Fixed bug #115 "QMsm::isInState() implementation missing"
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:
- Modified the QP/C++ ports to Windows (both Win32 API (Windows) and Win32-QV (Windows with 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.
- Added portable "safe" macros from
<string.h> to the QP/C++ ports to Windows. These macros encapsulate the differences between Microsoft Visual C++ and other compilers (such as MinGW).
- 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.
- Modified the QF::stop() function in the QP/C++ port to Win32-QV (Windows with QV), so that it unblocks the QV event-loop and thus lets the application terminate.
- Modified all examples for Windows to use the new port structure.
- 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.
- 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.
- 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)
- Improved the PC-Lint support for checking the application-level code located in in
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 QP/C++ Design shows the current class hierarchy.
- 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:
- 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.
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.
- 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.
- 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".
- Several README files have been updated.
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 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 Windows or 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 QP/C++ Design shows the current class hierarchy.
- 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:
- 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 source 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.
- 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).
- The QP/C++ Ports folder has been reorganized to contain all currently available QP/C++ ports. The ports are organized into three categories: native QP/C++ ports ("bare-metal"), ports to 3rd-party RTOSes, and ports to big operating systems (Windows and Linux).(NOTE: the ports are now documented in the this QP/C++ Reference Manual. Each port sub-directory contains a README link to the corresponding page in the online documentation)
- The QP/C++ Examples folder has been reorganized to reduce the repetitions and contains all currently available QP/C++ examples. The folder includes four categories of examples: native QP/C++ examples ("bare-metal"), examples for 3rd-party RTOSes, examples for big operating systems (Windows and Linux), and 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 examples 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 QP/C++ Reference Manual. Each example sub-directory contains a README link to the corresponding page in the online documentation)
- A new 3rd_party 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 3rd_party 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.
- This release also comes with the much expanded online QP/C++ Reference Manual, which is cross-linked with the ports and examples.
Changes in detail:
- Renamed the "Vanilla" scheduler to the QV cooperative kernel for symmetry with the QK preemptive kernel. Renamed QF::onIdle() callback to QV::onIdle().
- 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.
- 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.
- Updated 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.
- Added the native port and examples to the classic ARM7/9 with AT91SAM7S-EK board and the IAR-ARM toolset.
- Added the native port and examples to the AVR (AVRmega) with GNU-AVR and IAR-AVR toolsets. The examples are provided for the Arduino-UNO board.
- Added the native port and examples to MSP430 with TI CCS-430 and IAR-430 toolsets. The examples are provided for the MSP430 LauchPad boards (the MSP-EXP430G2 and MSP-EXP430F5529LP for the "classic" MSP430 and "extened" MSP430X, respectively).
- Added port to 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.
- Updated port to embOS. Examples are available for STM32 STM32F4-Discovery board with IAR-ARM toolset.
- Updated port to FreeRTOS for the latest version 8.2.1. Examples are available for TI EK-TM4C123GLX board with GNU-ARM and IAR-ARM toolsets.
- Added port to Thread-X. Example is available for the Thread-X demo with Visual Studio on Windows.
- Updated port to 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.
- Updated 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.
- Updated the QP-Qt Port (QP/C++ port to the Qt cross-platform GUI framework).
- Added new port to Win32-QV (Windows with cooperative "Vanilla" scheduler, previously known as Win32-1T).
- Updated the lwIP-QP example for EK-LM3S6965 board.
Version 5.3.1, 2014-09-20
- QP/C++ 5.3.1 remains backwards-compatible with all QP/C++ ports and applications
This release fixes the following bugs:
- QMsm::isInState() returns invalid result (bug #105)
- QF::gc() doc typo (bug #102)
- POSIX-port Makefile error (bug #65)
- 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: http://www.state-machine.com/qpcpp
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.
- 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:
- Moved detailed documentation of functions from the header files (.h) to implementation files (.cpp).
- Removed the header file "qevt.h" and merged its contents into "qep.h"
- Added macros: trace records QS_QEP_TRAN_HIST, QS_QEP_TRAN_EP, and QS_QEP_TRAN_XP to "qs.h"
- Added macros: Q_TRAN_HIST(), QM_TRAN_HIST(), QM_TRAN_EP(), QM_TRAN_XP(), and QM_SUPER_SUB() to "qep.h"
- Added attributes entryAction and initAction to the QMState struct in "qep.h" (needed for transition to history).
- Added attribute act to the QMAttr union in "qep.h" (needed for transitions to entry point in submachine states).
- Changed the QState return type from action/state handler functions to uint_fast8_t.
- Added QM models to the "Transition to History" design pattern example (directory qpcpp) and also added an example for "Transition to History" with the QMsm class (see qpcpp).
- Changed the m_prio attribute of QActive to uint_fast8_t.
- 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.
- Changed the type of the tickRate argument in QTimeEvt constructor and ::QF::tickX_(), and ::QF::noTimeEvtsActiveX() to uint_fast8_t.
- Changed the type of the poolSize argument in ::QF::poolInit() to uint_fast16_t.
- Changed arguments evtSize and margin in ::QF::newX_() to uint_fast16_t.
- Changed attribute bits in QPSet8 as well as bytes and bits in QPSet64 to uint_fast8_t.
- Changed the QEQueueCtr event queue counter type to uint_fast8_t.
- Changed type of arguments qLen and margin in ::QEQueue::init() and QEQueue/QActive::post() to uint_fast16_t.
- Changed the return type from QK_schedPrio_() (priority) as well as the p argument in QK_sched_() and QK_schedExt_() to uint_fast8_t
- 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.
- Updated all make scripts for QP/C++ ports to include the new qmsm_in.cpp in the QP/C++ library builds.
Version 5.2.1, 2014-01-06
This release fixes two bugs.
- 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.
- In file qs_mem.cpp fixed an error in accounting used bytes in the QS trace buffer.
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 http://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:
- 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.
- 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.
- 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.
- 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.
- Added the QMsmTst example application analogous to the QHsmTst to test the QMsm base class. The QMsmTst comes with the QM model qmsmtst.qm.
- 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.
- 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
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:
- In file qf_tick.cpp reversed the logic inside QF::noTimeEvtsActiveX()
- Modified free entry accounting (nFree) in the files: qeq_init.cpp, qeq_fifo.cpp, qeq_get.cpp, and qeq_lifo.cpp.
- Modified free entry accounting (nFree) in the files: qa_init.cpp, qa_fifo.cpp, qa_get_.cpp, and qa_lifo.cpp.
- Introduced new macro QF_CPU_SLEEP() in the ARM Cortex-M Vanilla ports.
- Changed Board Support Package files (bsp.cpp) in the ARM Cortex-M Vanilla examples.
- Modified the CMSIS-compliant startup code in all ARM Cortex-M Vanilla examples.
- 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
- 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.
Version 5.1.0, 2013-10-02
- QP/C++ 5.1.0 remains backwards-compatible with the existing QP/C 4.x applications, except the ARM Cortex-M applications need to adjust the interrupt priorities. Specifically, you need to set the interrupt priorities equal or lower than QF_AWARE_ISR_CMSIS_PRI 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).
- Added the new QMsm class to qep.h. Changed the inheritance tree by deriving QHsm and QFsm from the QMsm base class.
- 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.
- Added the new QMActive class to qf.h. Extended the inheritance tree to derive QMActive from QActive.
- 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.
- 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.
- Modified the QActive::defer() function to return the status of the defer operation (true==success), instead of asserting when the defer queue overflows.
- 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.
- 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%-cm\ 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.
- 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.
- 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
- 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).
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:
- Added ports and examples for ARM Cortex-M4F with the EK-LM4F120XL board (Stellaris Launchpad).
- 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.
- Updated all ARM Cortex-M ports and examples to the latest IAR EWARM 6.50 and Sourcery CodeBench 2012.09-85.
- Updated App Notes "QP and ARM Cortex-M with IAR" and "QP and ARM
Cortex-M with GNU".
- Updated the PC-Lint support files (include\lib-qpcpp.lnt, au-misra2.lnt) to the latest PC-Lint 9.00j.
- Updated the Application Note: "QP/C++ MISRA-C:2004 Compliance Matrix".
- Spell-checked the comments in all QP/C++ source files and removed several typos.
- Removed the Qt ports and examples from the QP/C++ Baseline Code and moved them to the separate QDK/C++-Qt.
- QP/C++ Version 4.5.04 preserves full compatibility with QM 2.2.03 and all existing QDKs for QP/C++ 4.5.xx.
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.
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:
- 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".
- Eliminated QP port to Win32 with one thread (Win32-1T).
- Consolidated all QP ports to POSIX OSs (Linux, Linux64, Mac_OSX) into a single port to POSIX and placed it in the directory posix.
- Renamed the port directory qt_1t\ to qt.
- Added event constructor to qevt.h (controlled by the configuration macro Q_EVT_CTOR).
- Added the conditional compilation with the macro Q_EVT_VIRTUAL around the virtual destructor in the QEvt class.
- 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.
- 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.
- 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-cortex\ directory).
- Modified all examples to link only one QP library.
- Added QM models to most examples and used the automatically generated code from the models instead of the hand-written code.
- Modified Qt ports to use the event constructors and modified examples for Qt to demonstrate this feature.
- 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".
- Added new macro QS_USR_DICTIONARY() to QS for providing symbolic names for user-defined trace records
- Added new macro QS_RESET() to QS for telling the QSPY application when the target resets. This allows QSPY to reset its internal state.
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:
- Added macro Q_UNHANDLED() and return value Q_RET_UNHANDLED in qep.h.
- Modified qhsm_dis.cpp to handle the Q_RET_UNHANDLED return value.
- Updated the QP/C MISRA-C++:2008 compliance matrix to include the new MISRA-compliant way of coding guard conditions.
- 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.
- 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.
- Corrected qfsm_dis.cpp, which did not generate QS trace records for entry and exit from non-hierarchical states.
- Updated the IAR ARM compiler used in the ARM Cortex-M examples to the latest version IAR EWARM 6.40.
- 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.
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:
- 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.
- 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
- 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.
- 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.
- 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.
- 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.
- Added the QHsm::state() and QFsm::state() accessor methods.
- 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.
- 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.
- 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.
- Added QP port to Qt_1t (Qt with one thread) and two example applications (DPP and PELICAN crossing).
- Added GNU compiler option -pthread to QP ports for POSIX with P-threads, including QP ports and examples for Linux and Mac OS X.
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.
- 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.
- 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.)
- Added the PC-Lint option files std.lnt and lib-qpcpp.lnt to the qcppp/include/ directory.
- Cleaned the whole QP/C code from lint comments. All PC-Lint options have been moved to PC-Lint option files.
- 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.
- 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.
- Added macros Q_STATE_CAST() and Q_EVENT_CAST() to qep.h to encapsulate deviation from MISRA-C++:2008 rule 5-2-7.
- 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.
- Updated ARM Cortex-M examples with the latest CMSIS v3.0, which complies with more MISRA-C:2004 rules.
- Added DPP examples for MISRA-C++:2008 compliant applications (for IAR-ARM and GNU-ARM).
- Updated ARM-Cortex-M3 port with GNU to the latest Sourcery CodeBench 2011.09-60.
- 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.
- 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",
Version 4.3.00, 2011-11-03
- 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.
- 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.
- Updated all QP ports to comply with the new critical section names and policies.
- Modified the ARM Cortex-M port qk_port.s to take advantage of the new structure of the QK scheduler.
- Upgraded the examples for ARM Cortex with IAR EWARM to the latest IAR EWARM version 6.30.
- Upgraded the examples for ARM Cortex with GNU (CodeSourcery) to the latest Sourcery CodeBench 2011.07-60.
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.
Version 4.2.02, 2011-09-08
- 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.
- 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.)
Version 4.2.01, 2011-08-13
- 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.
- Added QP port and examples for Mac OS X on 80x86.
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:
- Changed the QEvent base class (file qevent.h). The private member 'dynamic_' has been replaced by two members 'poolId_' and 'refCtr_'.
- 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.
- 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).
- Added explicit call to the event destructor in QF::gc() (file qf_gc.cpp).
- 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.
- Made algorithmic changes in the QF source code related to the change of storing the event pool-IDs and reference counters inside QEvent.
- Changed the default signal size (macro Q_SIGNAL_SIZE in the file qevent.h) from 1 to 2 bytes.
- 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.
- 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.
- Added new QF macros #QF_TICK, #QF_PUBLISH, and #QACTIVE_POST in file qf.h to provide sender of the events for software tracing.
- Added new QS macros (files qs.h and qs_dummy.h) for handling 64-bit integers.
- Added the functions QS::u64() and QS::u64_() and new source file qs_u64.cpp.
- Added the QS macro #QS_U32_HEX_T for hexadecimal formatting of integer numbers in the user-defined trace records.
- Added the new port linux64 for 64-bit Linux. Also added the corresponding examples for 64-bit Linux.
- 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.
- 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.
- 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.
- Upgraded the examples for ARM Cortex with IAR EWARM to the latest IAR EWARM 6.20.
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:
- 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.
- 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.
- 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.
- 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.
- 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).
- Upgraded the examples for ARM Cortex with CodeSourcery to the latest Sourcery G++ 2011.02-2.
Version 4.1.06, 2011-01-07
- Fixed a bug in the software tracing instrumentation in the function QF::remove_(), file qf_act.cpp.
- Made cosmetic improvements to the example QM models of the "Fly 'n' Shoot" game.
- Made improvements in make.bat files for building the examples for DOS/Open Watcom to run better in DosBox on Linux.
- Upgraded the examples for ARM Cortex with IAR to the latest IAR EWARM version 6.10.
- Upgraded the examples for ARM Cortex with CodeSourcery to the latest Sourcery G++ 2010.09-66.
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.
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.
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:
- Unified source code for ARM Cortex-M3 and the new ARM Cortex-M0 cores, including the code for the preemptive QK kernel.
- Compliance with the Cortex Microcontroller Software Interface Standard (CMSIS) in all ARM Cortex examples.
- 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.
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.
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 <qpcpp>/examples/80x86/dos/watcom/l/calc2 shows how to derive state machine classes with QP 4.x. The SLS example located in <qpcpp>/examples/80x86/dos/watcom/l/sls shows the implemenation of the new State-Local Storage state design pattern.
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.
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.
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.
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.
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.
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():
- 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 not 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 not need to lock the scheduler. 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 eliminates the need for the recursive 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.