Re: Retrieving 'secret' setting

[Colin Helliwell <colin helliwell ln-systems com>]

> On 07 September 2017 at 17:13 Colin Helliwell <colin helliwell ln-systems com> wrote:
>
> > On 07 September 2017 at 16:37 Dan Williams <dcbw redhat com> wrote:
..
> > client = NM.Client.new(None)
> > c =
> > client.get_connection_by_uuid(sys.argv[1])
> > secrets =
> > c.get_secrets(NM.SETTING_WIRELESS_SECURITY_SETTING_NAME)
> >
> > # this merges
> > the secrets into the existing connection
> > c.update_secrets(NM.SETTING_WIR
> > ELESS_SECURITY_SETTING_NAME, secrets)
> >
> > # write the full connection back to NM
> > c.save()
>
> Thanks, I'll work through that in due course, but just to step back a bit to something perhaps simpler 
> (which may help my understanding in general!) I've just been trying to modify some other gsm settings. All 
> is good until trying to write the changes to the connection file (/etc/NetworkManager/system-connections/):
> c.save() does write to the file (I see the timestamp is changed), but the setting itself is unchanged, 
> whereas
> c.commit_changes(True,None) *does* alter the setting in there.
>
> Just wondering therefore what save() is intended for, and/or whether it needs to be used in a different way?
> No problem to use commit_changes() instead, but save() doesn't seem to be doing what the docs imply: "Saves 
> the connection to disk if the connection has changes that have not yet been written to disk, or if the 
> connection has never been saved"


Ok, think I get it now - I was seeing 'commit' as higher level as 'save'. But in fact commit is just on the 
loaded connections, with the option to save; and 'save' only saves what's been committed? 
Mind you I always seem to get False from get_unsaved(), regardless. (Not too bothered as I won't be using 
that ultimately - I just noticed it during my experimentation)



> AND, just to get my next question in while I'm at it :) - What are the calls to do a down/reload/up on the 
> connection? (i.e. I'll have changed the gsm settings from incorrect apn etc to correct, and want pppd to be 
> kicked into a re-connect)

**
I still haven't sussed out how to do this....
(Not even sure of the class/api to be looking at)
**


---

An additional query about changing the gsm password. I'm basically able to do this, but I've noticed that it 
doesn't seem possible to *remove* the password. For example the username can be deleted [i.e. from the 
connection file] with      settings.set_property(NM.SETTING_GSM_USERNAME, None)

Password can be *changed* with a set_property() but if it's changed to None then the connection file - though 
it is rewritten - doesn't change. 

It can be emptied with     set_property(NM.SETTING_GSM_USERNAME, "")
this puts 'password=' into the file but then an error is thrown ("A password is required to connect to 
'GSM'"). Note - for clarity of permutations in this info, username is set.

A workaround seems to be to explicitly set it to an empty string:
  set_property(NM.SETTING_GSM_USERNAME, '""')

Not a major problem, but I wondered if it's a bit of a bug?  The same seems to occur doing it with a 'nmcli 
conn mod'
(I'm using the 1.8.2 tarball)