GNOME Bugzilla – Bug 729845
Clean up connection matching algorithms
Last modified: 2016-01-20 14:55:34 UTC
Thomas says... ---- For the record: I am doubting this algorithm to remove the properties from the settings if we match it and then check at the end, that no settings are left. This seems wrong to me because the settings properties are present in the dictionary, *iff* they differ from the default value. Why would we remove non-default properties and ensuring that we removed that all? These default values don't seem any special for check_p_match(). Anyway, if it works for now, covering all previous cases + the bug you are trying to fix, I am fine with it. In general, I think our algorithm for matching is wrong/insufficient, and we will have to revise/improve it later -- which means, we really have to write good test cases now! Another thing that might be improved later, is that nm_utils_match_connection() keeps the first @best_match it finds, instead improving later on, if if finds a still better one... or if there are two equal connections, we should prioritize least-recently-used... ---- So lets continue that discussion here and maybe make some fixes.
One issue I found is bug 732703 for bond options.
We're probably not going to do anything additional here unless somethign really, really important comes up.
closing this bug. There are not concrete issues that needs to be fixed here. And in the long term, connection assumption should anyway work entirely different so that we don't generate-and-match connections. See bug 746440 and https://mail.gnome.org/archives/networkmanager-list/2015-November/msg00031.html