New release: v2.0.2

v2.0.2 has just been released. This is a minor release containing only bugfixes and cleanups. No new functionality has been added, but some functionality has been made optional to reduce RAM and ROM requirements on resource-constrained platforms. And no behavior has changed, except for platforms where CHAR_BIT != 8.

Reduced RAM and ROM requirements

Two new configure/preprocessor options have been added to reduce the runtime size of the object dictionary:

• --disable-obj-default (LELY_NO_CO_OBJ_DEFAULT): disable default values in the object dictionary.
• --disable-obj-upload (LELY_NO_CO_OBJ_UPLOAD): disable custom upload indication functions in the object dictionary.

Together with the existing options to disable names and minimum/maximum values, these options are no also taken into account when generating the static device description. This can significantly reduce the ROM size.

CiA 301 requires that upon receipt of the NMT “Reset application” command, the parameters of the manufacturer-specific profile area (object 2000…5FFF) and of the standardized device profile area (object 6000..9FFF) are set to their power-on values. The stack implements this requirement by storing the values of those objects as a concise DCF in memory on initialization, and restoring them when the NMT command is received.

However, embedded devices usually reboot in response to a “Reset application” command, which already resets the object dictionary. In that case, storing the concise DCF is not necessary and can be disabled with the --disable-dcf-restore configure option or the LELY_NO_CO_DCF_RESTORE preprocessor macro.

CHAR_BIT != 8

Nearly all computer architectures provide 8-bit bytes. However, one of the platforms we would like to support, the C28x Piccolo DSP (part of the C2000 microcontroller family of Texas Instruments), has 16-bit bytes. This is now supported by the CANopen stack and supporting libraries.

Byte arrays, such as CAN frame payloads and serialized values, are treated as sequences of 8-bit bytes, even if CHAR_BIT != 8. The most significat bits are discarded, if necessary. This wastes some memory on platforms like the C28x Piccolo DSP, but it keeps the code and type definitions simpler and more maintainable. To prevent mistakes, this choice has been made explicit in the function prototypes of the byte-order functions in <lely/util/endian.h>.

v2.0.2 still assumes the availability of the uint8_t typedef. This has already been rectified in the master branch, and from v2.1 onward, the stack will no longer assume the availability of any exact-width integer type. However, this fix touches too much code to be backported to v2.0.

Download

You can download the source from GitLab or the Ubuntu packages from our PPA.