On Wed, Dec 14, 2016 at 10:46:20 -0500, Laine Stump wrote: > On 12/13/2016 08:52 AM, Jiri Denemark wrote: > > Iterating over all child nodes when we only support one instance of each > > child is pretty weired. And it would even cause memory leaks if more > > than one <tftp> element was specified. > > ACK, but could you also look for dhcp[2]/tftp[2] and log an error if found? > I know there are *lots* of places we ignore extra elements in the XML, but > in this case there would be a silent behavior change if someone had > (erroneous) multiple tftp or dhcp elements - previously we would have > honored the final occurence of each element, but now we honor the first. So > even though it's their own fault, it would be nice to Well, if a user provided such a wrong XML, libvirt didn't do what the user asked for anyway. I don't think implementing a different way of not doing what they asked for is something we should worry about :-) That said, adding the check there is trivial, but I'm not sure it's a good idea since we do not check for extra elements anywhere else. > (BTW, I've always disliked that some of our XML parsing code iterates > through the raw nodes like this used to, and some uses XPath to get specific > nodes. It would be so much easier for newcomers to understand if we picked > on method and used it consistently...) Yes, I think we'll eventually rewrite all parsing to use XPath, but it will take some time. Jirka -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
