On Sun, Oct 09, 2016 at 10:14:10PM +0300, Ido Schimmel wrote:
Hi,
On Fri, Oct 07, 2016 at 03:10:31PM +0200, Ondrej Lichtner wrote:
> Hi,
>
> I've sent one final patch to the list that fixes the #28 regression
> test and continued with the v12 release.
>
> Right now I have the commits only locally so that I can fix stuff, and
> I've made scratch builds for rawhide (also usable by F23, F24, RHEL7)
> and EPEL6:
>
http://koji.fedoraproject.org/koji/taskinfo?taskID=15981799
>
http://koji.fedoraproject.org/koji/taskinfo?taskID=15984998
>
> Please try them out and send me any issues you find.
One of the machines I use it too old for these builds (fc22), but I
tried the git version (8e7117c8c5361920cc6b41b513bfa58fce2ec4c4) and
it seems that commit 82f606fb169e ("{Net, Nm}ConfigDevice: split up()
into up+address_setup") introduced a regression, which caused most of
recipes/switchdev to fail.
This commit breaks compatibility with older NetworkManager versions so
yes, problems can arise... it would be pretty hard to implement in such
a way that both versions are supported, which I don't think is important
enough since F23 is also reaching End Of Life (sometime in December-New
Year).
On that note... F22 is also not supported anymore, so it won't be
receiving the package update anyway, I suggest you update to a newer
version if possible :)
When reset() is called from within a recipe the up() method would
setup any addresses configured on the interface, but now it merely sets
the admin state. I added address_setup() just after it and it seems to
solve the problem.
Fair point, forgot about the InterfaceAPI calls... the patch looks good
so I'll push it today...
I'll do some more testing, look for more instances where this may occur
and send a patch.
Thanks for testing.
-Ondrej