merges with https://bitbucket.org/JaFojtik/canfestival-3

ID: 3459307 - Async problem in drivers/win32/win32.c

Windows CAN DLL driver sanity check.

Multiple instances CAN wirtual network for PEAK and Windows.

There is no English word "transfert". Fix all the typos.

Applyed changes from discussion '[Canfestival-devel] some minor CanFestival build suggestions', https://sourceforge.net/mailarchive/message.php?msg_id=28397766

[ canfestival-Bugs-3458203 ] Objdictedit

[ canfestival-Patches-3481330 ] Useless structure SHORT_CAN
[ canfestival-Patches-3481339 ] Useless local variable res in CanFestival\drivers\win32.c

Merge further Ica changes

Win32 IXXAT driver:
CHANGED: - Removed deprecated call to VCI_GetBrdTypeByName().
- The board type given to VCI2_PrepareBoard() is now always 0, because this parameter is not used.
- Removed the board_name parameter from IXXAT::open(). The board that should be opened is identified by the parameter board_number.
The correct parameter board_number can be determined using XAT_EnumHwEntry() and XAT_GetConfig().
- Changed the constructor IXXAT::IXXAT(). board->busname now only contains the board_number, see XAT_EnumHwEntry() and XAT_GetConfig().

Win32 IXXAT::
CHANGED: - removed newline from log messages

pdo.c:
FIXED: - fixed two wrong MSG_ERR calls

Win32:
CHANGED: - Switch character encoding to unicode.
- Changed logging format strings from %s to %S.
FIXED: - second argument to swprintf() is now the size of the target buffer

Win32 IXXAT:
CHANGED: - Changed logging format strings from %s to %S.

Win32:
FIXED: - The argument of LoadCanDriver() is LPCTSTR, not const char*

Win32 IXXAT driver:
ADDED: - Two additional log messages added, when initialization of the driver fails.
CHANGED: - Instead of writing directly into OutputDebugString() messages are printed using the new macro MSG_ERR_DRV().

Win32 applconfig:
CHANGED: - The MSG()-Makros for Visual Studio both print now to OutputDebugString().
ADDED: - Makro CANFESTIVAL_DEBUG_DRV_MSG() and MSG_ERR_DRV() to print into MSG() from a can driver implementation.

Win32 IXXAT driver:
FIXED: - A return value of 0 from VCI_ResetCan() and VCI_StartCan() indicates an error, too.
CHANGED: - The return values of VCI_ResetCan() and VCI_StartCan() are logged in case of an error.

SDO transfer:
FIXED: - Wrong SDO line could get changed due to a wrong parameter to setSDOlineRestBytes().

Win32-Dll:
CHANGED: - export the function sendOnePDOevent from the windows dll.

Win32-Timer:
CHANGED: - Timer implementation for win32 from absolute time (_ftime) to GetTickCount().
This fixes the following bug: When the system time was changed, a heartbeat timeout occurred.
FIXED: - Timeout seems only work properly if EnterMutex() is called before call of GetTickCount() (Patch from Roland Marquis)

FIXED: - The AVR example now compiles in linux.
- adapted AVR code to current default branch.

FIXED: - adapted paths for current vscom driver
- removed chmod +x vs_can_api.dll;
- added option -o to unzip command to overwrite existing files

FIXED: - the macro SUB_PROG_CFLAGS was overwritten for 64bit machines using gcc. Because of this the xenomai compilation failed.

Merged some canfestival-3-ica Fixes.

FIXED: - The AVR example now compiles in linux.
- adapted AVR code to current default branch.

FIXED: - adapted paths for current vscom driver
- removed chmod +x vs_can_api.dll;
- added option -o to unzip command to overwrite existing files

FIXED: - the macro SUB_PROG_CFLAGS was overwritten for 64bit machines using gcc. Because of this the xenomai compilation failed.

bug correction in sdo.c, in writeNetworkDictCallBackAI a call to _writeNetworkDict had endianize forced to 1

bug correction in dcf.c, an entry of only one byte was not checked

New feature : event timers allowed for RxPDOs (subidx 5 in RxPDO com paramaters)

Configuration manager with DCF in object 0x1F22 rewritten almost from scratch.
1. The boot-up message from a device starts a verification of entries
2. If all entries matches the node is started
3. If an entry differs the whole dcf is written and a save is done
4. A reset is send to the node
5. If several boot-up are received at the same time they will be managed one
by one thus only one free sdo client is needed for the whole process.

Added some fixes from Jaroslav Fojtik submitted on sourceforge bug tracker:
3096348 Missing "break" in states.c/canDispatch()
3096359 Crash inside dcf.c/send_consise_dcf()
3101891 Potencial overflow inside getNodeState()

Added include stdlib.h for malloc functions in sdo.c

patch from Stefan Kratochwil <entwicklung@inovel.de> : canfestival-3-fm3_698.patch
Bug:
If an object dictionary entry was requested whose size exceeds
SDO_MAX_LENGTH_TRANSFERT, the memcpy() call at line 139 of objacces.c overwrites
the memory after *pDestData which causes stack corruption.
-> Bugfix:
The bug was corrected by size checking the requested data. An 'Out of
memory' error message will be generated if the requested data exceeds
SDO_MAX_LENGTH_TRANSFERT.

Additional changes:
Added dynamic buffer allocation for the SDO transfer. This feature can be used
if SDO_DYNAMIC_BUFFER_ALLOCATION is defined in config.h. The size of the
dynamically allocated buffer is controlled with
SDO_DYNAMIC_BUFFER_ALLOCATION_SIZE.
-> Note:
This change removes the detection of OD_LENGTH_DATA_INVALID errors!

patch from Christian Taedcke <hacking@taedcke.com> : fix-wrong-resetSDOline-for-sdo-timeout.patch
FIXED: - The sdo line must not be closed at the end of SDOTimeoutAlarm() if the same sdo line is closed and reopened for a new transfer inside the callback. (Patch from Jaroslav Fojtik)

CHANGE: SDO block mode added, without CRC support
WARNING : API change in client functions

timers_unix.c : remove sigint and sigterm catch
sdo : Allow multiple servers
The sdo transfer struct is not anymore referenced by server's node id but by
client or server number in the OD. Node id is not relevant in SDO transfert.

fixed bug with cross wmingw build

Removed unneccessary overhead introduced with the last fix.

Fixed incorrect struct access in sendPDO() and inserted a missing variable in sendOnePDOEvent(). These errors occured at compile-time when configuring the project with './configure --debug=PDO' or './configure --debug=WAR'.

Fix some side effects of EDS import, while editing afterward

FIXED: - changed make targets from "%o: %c" to "%.o: %.c" (Patch from irc <oleg_osov>)
find . -name 'Make*' | xargs perl -w -i.bak -p -e "s/\%o: \%c/%.o: %.c/g"

FIX: - if a sdo transfer timeout occurres, reset the sdo line even if the callback function does not.

FIXED: - Sdo lines with the internal state SDO_ABORTED_ITERNAL are now closed.

CHANGED: - canSend message parameter is now a const pointer
ADDED : - IXXAT specific can bus watchdog to restart the VCI2 driver if a bus off, data ovverrun or remote queue overrun is detected. The can adapter cannot recover on its own.
* * *
CHANGED: - IXXAT::m_watchdogTimerId from DWORD to UINT_PTR to fix compiler warning

FIXED: - compiler warning from printf format string

FIXED: - TimerInit() was missing

FIXED: - OK and NOT OK was wrong in canSend()

FIXED: - bug while normal sdo transfer initialization (setting the byte count)

FIXED: - Renamed macro for peak win32 driver, patch from mailinglist (post from Christian Boepple, 2010-09-08 14:38)

CHANGED: - function sendOnePDOevent (CO_Data* d, UNS32 pdoNum) into sendOnePDOevent (CO_Data* d, UNS8 pdoNum)
- added explicit casts to remove compiler warnings

CHANGED: - added explicit cast to remove compiler warning

CHANGED: - canReceive and canSend return now 1 on error and 0 on success

CHANGED: - added explicit cast to remove compiler warning

FIXED: - added CALLBACK to can callback function prototypes

FIXED: - added missing endif
* * *
FIXED: - moved #include <linux/delay.h> into the ifdef __KERNEL__ block

CHANGED: - async access queue has now blocking read