On Wed, Nov 24, 2010 at 7:04 AM, Dan Williams <d...@redhat.com> wrote: > On Sat, 2010-11-20 at 15:48 +0300, Andrey Borzenkov wrote: >> 1) if plugin fails to write connection to stable store (for whatever >> reasons) NM is left with old, cached, connection that was received fro >> client. It makes impression that connection update worked (and it even >> will work for as long as either NM is restarted or connection is >> re-read again from disk). The reason is, >> src/system-settings/nm-sysconfig-connection.c:pk_update_cb() first >> updates in memory and then calls to save new definition. It is even >> documented in comments :) >> >> /* Update our settings internally so the update() call will save the >> new >> * ones. >> >> But if ->update fails, settings are not rolled back. >> >> It could be possible; we could add old_connection field to PolkitCall >> and use it to rollback in con_update_cb. Any potential issues >> associated with it? > > Yeah, we could do that, and hopefully return an error to the client that > called update() in the first place. >
Error is already returned; patch for review is sent in separate message. >> 2) NM core does send proper error via D-Bus to nm-connection-editor, >> but it simply does not have code to display them. On update it finally >> calls connection_updated_cb() and it does not display anything (nor is >> any error displayed on callback chain). > > Yeah, we should do better here. I thought we had an error dialog that > popped up for stuff like this? The code there is somewhat convoluted, > but ideally we'd notify the user that an error occurred. > Unfortunately this is beyond my knowledge :) _______________________________________________ networkmanager-list mailing list networkmanager-list@gnome.org http://mail.gnome.org/mailman/listinfo/networkmanager-list