msxml3/tests: add additional conformance tests that show incorrect xpath behavior in wine

[Nikolay Sivov]

On 6/29/2013 20:26, John Chadwick wrote:
> On Sat, Jun 29, 2013 at 4:48 AM, Nikolay Sivov <bunglehead at gmail.com> wrote:
>> On 6/27/2013 11:46, John Chadwick wrote:
>>
>>> +    /* msxml's selectNodes returns a document ordered nodelist,
>>> regardless of
>>> +     * whether or not the xpath nodelist was document ordered... */
>>> +
>>> ole_check(IXMLDOMNode_selectNodes(rootNode,_bstr_("ancestor-or-self::node()"),
>>> &list));
>>> +    todo_expect_list_and_release(list, "D1 E2.D1");
>> Please use explicit check for return value, like ok(hr == S_OK, "...");,
> Which return value am I missing? FWIW, I'm doing what the other tests
> are doing, as far as I can tell.
What I mean is that use of macros like ole_check() is discouraged, it's 
better to have explicit check for return value.
>
>> also you probably don't need another macro for list checking,
>> isn't it enough to wrap expect_list_and_release() in todo_wine {} ?
> Could it be that simple? I tried to do
>> todo_wine expect_list_and_release(list, "D1 E2.D1");
> But it fails before the actual test occurs, because another test
> inside the macro succeeds, apparently. Are you saying if I wrap the
> expect_list_and_release macro call in a scope, it will work? I will
> try.
Yes, I see what you mean. In order to improve situation I'd suggest to 
move to using
table of test queries and results. This way you have less code and no 
duplication,
todo flag should just be stored in test data. We already have some tests 
done this way,
it's much easier to maintain.
>
> I figured creating a new macro would be a better solution than
> duplicating the code separately in two locations, but if it'd be
> better to just not use the macro, then that's simple enough to fix.
>