On Thu, Mar 14, 2013 at 12:22 PM, Jukka Rissanen
<jukka.rissa...@gmail.com> wrote:
> Hi Marcel,
>
> On 14 March 2013 19:18, Marcel Holtmann <mar...@holtmann.org> wrote:
>>>> In general we want to be able to enable/disable debug messages at runtime. 
>>>> That is where I want this to go. Same as the kernel allows to 
>>>> enable/disable dynamic debug at runtime.
>>>
>>> Sure, why not. My patch does not contradict with this goal.
>>
>> I like to have one way and one way only. And actually "connmanctl debug 
>> <pattern>" sounds like a pretty good way of giving this to a user.
>>
>
> Hmm, do you mean that the user would not be able to activate debugging
> via -d option any longer?

I don't see a reason to remove the cmdline option, it does serve a
purpose to enable debugging right from the start. Or even just running
it in the foreground in a terminal.

But from a user/developer point of view, we should not ship a service
file that jumps through hoops to enable/disable debugging - that is
something a developer locally should do and they should know how to.

Having the `connmanctl debug <bool>` option will allow you to pass
instructions like this:

- disconnect your service
- run `journalctl /usr/sbin/connmand` in a terminal to start capturing
debug output
- type `connmanctl debug on`
- connect your service

That seems really easy and friendly to me.

There's also an alternative to modifying the connman.service file.
With systemd v198 you can now do:

mkdir -p /etc/systemd/system/connmand.service.d
echo -e "[Service]\nExecStart=/usr/sbin/connmand -n -d\n" >
/etc/systemd/system/connmand.service.d/debug.conf
... daemon-reload
... restart

I didn't test this myself though. And it still looks a bit like a kludge.

Auke
_______________________________________________
connman mailing list
connman@connman.net
http://lists.connman.net/listinfo/connman

Reply via email to