Possible bug between Sugar and NetworkManager F11

[James Cameron]

On Thu, Sep 02, 2010 at 09:29:03AM -0300, Franco Miceli wrote:
> While making mods to NM autoconnection feature I've come across a
> weird behaviour from the favourites connections
> (/home/olpc/.sugar/connections.cfg) reported to NetworkManager at
> startup.
> 
> What I've seen is that -only at startup- just one connection gets
> reported to real_get_best_autoconnection (a function within
> nm-device-wifi.c), thus being this one the one that gets choosen by NM
> (if there's an AP that matches it). At first I thought it was that the
> wireless scans did not report all the APs within range, but later -by
> debuging the program I found out that only one favourite connection
> was reported (from several in connections.cfg) and all APs where seen.
> This behaviour doesn't repeat itserf if the autoconnection feature
> gets executed after this first time.

Strange.  But worth fixing.  I doubt it is correct.

> Is this something done on purpose for some reason? Is sugar
> responsible for this or is NM? If it is sugar, should I open a ticket
> or could someone point out where in jarabe/... can I find this first
> connection report?

I can't see in the code where this might be happening, but I can tell
you where to look:

src/jarabe/model/network.py:

class NMSettings

NewConnection is a D-Bus signal that is sent by Sugar to NetworkManager
for each connection read from connections.cfg.

NewConnection is triggered by NMSettings.add_connection method.

NMSettings.add_connection is called by the function add_connection.

add_connection is called by load_wifi_connections for each entry in the
connections.cfg file.

p.s. you sent back to the mailing list the whole digest message you
received.

-- 
James Cameron
http://quozl.linux.org.au/