Re: device removal signals does not work for me

[Thomas Bechtold <thomasbechtold jpberlin de>]

Hi Jirka,

thank you very much for the description and the fix. is it possible to 
fix this in the 0.8 branch, too?

Cheers,

Tom

On 24/11/11 17:57, Jirka Klimes wrote:
> On Wednesday 23 of November 2011 07:33:07 Thomas Bechtold wrote:
> > Hi Dan,
> >
> > any other idea why the example code does not work?
>
> I've look into the problem. It showed up that it is really libnm-glib problem.
>
> DeviceAdded and DeviceRemoved signals are seen in dbus-monitor output. And
> they are also delivered to libnm-glib.
>
> However, what was the problem is that NMClient/NMDevice in libnm-glib use lazy
> approach and get data from NM over D-Bus only after they are asked for them
> (or get them via signals). This is nice in general. However, in case of device
> removal, libnm-glib tries to get devices from NM only after receiving
> DeviceRemoved signal. But at this point the device is not in NM any more, so
> no info is available about it (apart object path received as DeviceRemoved
> signal argument).
>
> So, when device is added *before* NMClient runs, NMClient won't have any info
> about that. When the device is added *while* NMClient is already running,
> libnm-glib will request devices upon DeviceAadded signal and thus the data
> will be available later upon DeviceRemoved signal.
>
> I pushed fixes to libnm-glib to correct the behaviour. Now, the device data
> are obtained from NM in constructor.
> 42060fdd897749c79fcc60ac16bb6fa67e4c8fd3
> 576acdd2bf4cfde11f9b8ea15ec87bb7271c5502
>
> I hope the explanation is understandable.
>
> Jirka
>
> PS:
> Dan, one question for you:
> NM's NMDevice has ip4-address property, which is not in libnm-glib. So I have
> to use a little hack (removing the propery from GetAll()'s reply), so that
> NMObject doesn't print errors.
> Do you think we should remove "ip4-address" from NM's NMDevice? In my opinion,
> "ip4-config" is sufficient and "ip4-address" is redundant now.