Re: [libvirt] [RFC] conf: Order of AddImplicitControllers and DomainDefPostParse
On 06/14/2013 02:34 PM, Viktor Mihajlovski wrote:
Implicit controllers may be dependent on device definitions altered
in a post-parse callback. E.g., if a console device is
defined without the target type, the type will be set in QEMU's
callback. In the case of s390, this is virtio, which requires
an implicit virtio-serial controller.
By moving the implicit controller definition after the post-parse
procssing this can be fixed. OTOH the implicit controllers might
need post-parse actions as well, although I am currentyl not
aware of one.
What would speak against swapping the order as suggested in
the patch below? Feedback welcome.
I can't say I'd be against that. I hope there's no implicit controller
that needs post-parse function, because when there is, all the actions
can be taken care of in the function. And if same things need to be
checked for, deduplication of such code takes care of it. In case such
code movements will show that these two functions would be better off
merged together, than it can be acted upon (I'm going to extremes with
this unnecessary explanation).
However, it would be nice if you could also add a test case to show and
check for the behavior you're trying to achieve.
Other than that I'd be ok with such movement, but also happy if someone
else provided his/her opinion as well.
Have a nice day,
Martin