On 08/04/2011 07:12 PM, Robin Holt wrote:
> On Thu, Aug 04, 2011 at 07:01:40PM +0200, Wolfgang Grandegger wrote:
>> On 08/04/2011 06:41 PM, Robin Holt wrote:
>>> On Thu, Aug 04, 2011 at 06:35:06PM +0200, Wolfgang Grandegger wrote:
>>>> Hi Robin,
>>>>
>>>> On 08/04/2011 06:23 PM, Robin Holt wrote:
>>>>> Add a wrapper function for a register dump when CONFIG_CAN_DEBUG_DEVICES
>>>>> is set.
>>>>>
>>>>> Signed-off-by: Robin Holt <[email protected]>
>>>>> To: Marc Kleine-Budde <[email protected]>
>>>>> Cc: [email protected]
>>>>
>>>> This patch is useful for development but should be dropped for mainline.
>>>
>>> I am not sure why?  There is already a CONFIG_CAN_DEBUG_DEVICES Kconfig
>>> setting.  This just makes it of use for flexcan.c.  The code is not
>>> made less readable by the patch and when CONFIG_CAN_DEBUG_DEVICES=n,
>>> there is no code difference with and without the patch.  It seems like
>>> we would minimize bit rot if it were maintained in the mainline tree.
>>> What are the arguments against having it included?
>>
>> Debugging output should be reduced to a few useful messages. A normal
>> user will not understand the register content of that controller and
>> may users use CONFIG_CAN_DEBUG_DEVICES=y by default. Though there should
>> be no debug output in the TX and RX patch.
> 
> I did not know that about CONFIG_CAN_DEBUG_DEVICES.  What if I change
> the conditional to something more like:
> 
> #if defined(CONFIG_CAN_DEBUG_DEVICES) && defined(FLEXCAN_DEBUG).
> Then a developer would have to get FLEXCAN_DEBUG defined, but normal
> users with CONFIG_CAN_DEBUG_DEVICES=y would not be affected.  Would that
> be acceptable?

Fine for development but not for mainline. None of the other drivers
does have advanced debugging capabilities.

Wolfgang.
_______________________________________________
Socketcan-users mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/socketcan-users

Reply via email to