peak-can-driver: update to 8.7.0

[sashko]

Updates:

• Fixed potential NULL pointer access bug when initializing in FD mode a non FD hardware
• PCANFD_TYPE_ERROR_MSG are now posed only if something new has to be notified
• BUS_LOAD notifications are now posted when bus_load has really changed
• Fixed some rare negative time values in timestamps
• Fixed old API init settings conversion ("echo i > /dev/pcanxx")
• Init settings are no more set to their defaults when the device is closed
• Fixed /proc/pcan "btr" column and sysfs "btr0btr1" node values
• Fixed user bittimings and conversion
• Changed mutex into spinlock when browsing the global devices list
• Be more permissive with PCANFD_TYPE_CANFD_MSG written on a non-FD device if data_len is <= 8
• src/pcan_fops.c: fixed double mutex_unlock() call
• src/pcanfd_core.c: fixed potential task deadlock waiting for the event to be signaled
• usb: reworked complete release of USB ressources
• usb: fixed wrong loop condition in release loop when probing a channel fails
• usb: PCAN-USB Pro and PCAN-USB hardware timestamsps are correctly handled now
• usb: fixed PCAN-USB X6 channel_number returned by PCANFD_GET_STATE
• xenomai: support of Xenomai 3 with full access to /sys/class/pcan
• xenomai: fixed mininal_xenomai dir RT applications for Xenomai 3
• xenomai: test/Makefile: added --no-auto-init flag to xeno-config
• xenomai: driver/Makefile: changed "make xeno3" into "make xeno"
• xenomai: pcanfdtst.c: added rendez-vous semaphore based to properly exit from main task (^C no more needed)
• rtai: support of RTAI 5.1 with full access to /sys/class/pcan
• rtai: added "minimal_rtai" subfolder in the "test" folder which is the same than "minimal_xenomai": two very small applications that read and write a bus
• lib/Makefile: RT: disabled 32-bit compilation with 64-bit arch
• pcanfdtst: fixed wrong help text of -B --brs option
• receivetest: added 'x' flags for STATUS msgs received
• test/*.cpp: fixed some bad indentations that give warning with gcc 6.3.0
• netdev: incr netdev stats overrun and error counters in case of DATA_OVR
• netdev: fixed pcan tx buffer overflow in case of flooding from the application
• netdev: fixed a NULL pointer dereference in DEBUG mode msg
• netdev: useless (chardev) rx FIFO object is now undefined in this mode
• handle single-shot mode for CAN FD devices
• netdev: fixed potential NULL pointer accces when interface is UP and the corresponding device is initialized (again) through the chardev interface
• netdev: fixed potential NULL pointer access with PCAN-PCI Express FD canX interface running in non-MSI mode
• netdev: removed useless "last_tx" init (this fixes lots of compilation issues)
• netdev: don't normalize bittimings since they already are. This prevents from changing them to another n-uple

Signed-off-by: Oleksandr Kravchuk oleksandr.kravchuk@pelagicore.com

[sashko]

Please write why there is a need to update the version, compatibility-reasons, what (new) features that can be expected, etc

No need. It's simply nice to have the latest fixes, especially when an update is a trivial recipe bump.

[martin-ejdestig]

It may be a bit superfluous for "pure" version updates like this, but... on the other hand... if you just put a "upd to latest to stay in sync" or something at the top before the shortlog, your intention is clear to reviewer... and posterity. Simple answer to "why?". :)

[thenor]

Why are the changes between versions only in the PR and not in the commit message? As already suggested, at least add something about the reason, even if I would prefer to have all the commits from the layer in the commit message. It makes for long commit messages, but I prefer that over not seeing what changed between the versions.