Mon, Aug 07, 2017 at 11:41:17AM CEST, olichtne(a)redhat.com wrote:
On Mon, Aug 07, 2017 at 11:34:14AM +0200, Jiri Pirko wrote:
> Mon, Aug 07, 2017 at 11:21:31AM CEST, olichtne(a)redhat.com wrote:
> >On Mon, Aug 07, 2017 at 11:02:40AM +0200, Jiri Pirko wrote:
> >> Mon, Aug 07, 2017 at 10:52:02AM CEST, roid(a)mellanox.com wrote:
> >> >
> >> >
> >> >On 07/08/2017 11:45, Roi Dayan wrote:
> >> >>
> >> >>
> >> >> On 07/08/2017 08:19, Jiri Pirko wrote:
> >> >> > Mon, Aug 07, 2017 at 06:17:45AM CEST, roid(a)mellanox.com
wrote:
> >> >> > >
> >> >> > >
> >> >> > > On 06/08/2017 11:17, Roi Dayan wrote:
> >> >> > >
> >> >> > >
> >> >> > > another method I was thinking is to make Iperf3 test
module to inherit
> >> >> > > from Iperf and set the binary in an attribute for the
compose cmd.
> >> >> >
> >> >> > I like that approach better.
> >> >> >
> >> >>
> >> >> was just thinking. I should not really inherit from Iperf as
it's
> >> >> another test module that might not exists if i didn't sync it
with
> >> >> sync_resources() ?
> >> >>
> >> >
> >> >so tested and I cannot do Iperf3 to inherit from Iperf as
> >> >Iperf might not exists if not synced.
> >>
> >> Can't they be both in the same file?
> >>
> >> Ondrej?
> >
> >I'm afraid not... the old lnst test module mechanism relies on the fact
> >that the python file has the same name as the test module class it
> >defines.
> >
> >I guess you could potentially split it into 3 test module files/classes:
> >IperfCommon, Iperf2, Iperf3 so you could save at least some code
> >redudancy, but then you'd have to always require syncing of the Common
> >class manually and even then I'm not sure if the sync mechanism could
> >handle this (since file names and locations change and this isn't
> >handled in old lnst).
> >
> >So for now... the easiest option is to just create a completely new
> >Iperf3 module that is independent.
>
> We can do it for now.
>
> Will the new syncing you are working on wrt pyrecipes handle this nicer?
>
> Thanks!
Yes, the test classes as they are now implemented already support this
to a certain extent, still need to finish some stuff wrt defining custom
classes that aren't in a well defined lnst location but other than that
it should work implicitly (automatically when an object is sent to the
slave) and it should also resolve the class dependencies.
Good. Looking forward to it!