Re: [libvirt] [PATCH v2 1/5] Internal refactory of data structures

On 07/18/2012 10:31 AM, Michal Privoznik wrote: Absolutely add a testcase if you are enhancing the xml parser to accept new tags. If you're going to touch docs/schemas, then it would also be nice to touch the counterpart docs/*.html.in. Unfortunately, I think our capability.rng is currently underdocumented in the html. Hmm, this says that: <secmodel> <model>...</model> <doi>...</doi> <model>...</model> <doi>...</doi> </secmodel> will validate, but: <secmodel> <doi>...</doi> <model>...</model> <model>...</model> <doi>...</doi> </secmodel> will not. I think that's somewhat good (since the parameters are positional, and we insist that they come in pairs, then the pairs must be properly interleaved), but did the C code enforce that? Indeed - that makes the XML layout problem MUCH easier. If we add an <interleave> around the 'model' and 'doi' sub-elements, then you could write: <secmodel> <doi>...</doi> <model>...</model> </secmodel> <secmodel> <model>...</model> <doi>...</doi> </secmodel> and have an unambiguous 2-model design, without quite so much attention being placed on the exactly-one 'doi' and 'model' subelement per 'secmodel'. -- Eric Blake eblake(a)redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org