On 26.08.2015 17:53, Laine Stump wrote:
<snip/>
How about this: give virNetlinkDelLink() a new argument
"fallbackDeleteFunc" which would either be NULL, or a pointer to
virNetDevBridgeDeleteWithIoctl() (in this case)? That avoids needing to
split virNetlinkDelLink into multiple parts as well as the duplication
of code, and I think it's much less obtuse than the boolean thing of V1.
D'oh! I mean, if someone calls virNetlink*() he can assume that the
function will talk to kernel via netlink. This would break the
assumption. But I because I am unable to come up with a better idea, I
think it's our best option.
Michal