Re: [Libvirt-cim] [PATCH 00/20] REWORK/PARIAL: Changes to solve unsupported tag issue

On 11/20/2013 03:36 AM, Xu Wang wrote: I hope the issue Viktor ran into with running out of memory shows that a duplicitous design is problematic. If you have to - rather than strdup() value, set INACTIVE, and return (unchecked) strdup()'d memory - it would have been just as easy to return the memory already allocated, set tmp->value = NULL, and set INACTIVE. That way you know it's "managed" by something else. The free(NULL) is a no-op anyway... Be sure to use/change the "xml_parse_test" in order to validate. In fact that print function you created could be called by xml_parse_test - at least temporarily. At first everything would be unknown. As you update parse functions, you should be able to prove that you've gone from unknown to known... Eventually your known should be the same output as what one could get 'today' if they ran the test (I forget the exact switches, but you'll get to know them). I think 2 & 4/5 need to be intermingled. That is when you update 'parse_fs_device', also update 'disk_fs_xml'. I'd change 'parse_block_device', 'disk_block_xml', and 'disk_file_xml' all at once. Change all the network functions together. Etc. I'm not exactly sure what this means, but it seems like this would be after step 5. Although it cannot be ignored or forgotten when devising the mechanism to parse/generate XML The change validity concern when there's unknown What's the difference between 4 & 5? Be conservative - I know it's hard to not make all the changes at once, but let's get the infrastructure right first. Updating the various _parse() and _xml() modules is a rote operation, but splatting out 50 patches at a time is tough to review/test... John It's the "unmanaged" or "unrealized" list... John