advapi32/tests: Test for last error after RegQueryValueEx

[Nikolay Sivov]

On 2/26/2010 19:00, Alexandre Julliard wrote:
> Nikolay Sivov<bunglehead at gmail.com>  writes:
>
>    
>> On 2/26/2010 16:38, Alexandre Julliard wrote:
>>      
>>> Nikolay Sivov<nsivov at codeweavers.com>   writes:
>>>
>>>
>>>        
>>>> @@ -626,15 +627,22 @@ static void test_query_value_ex(void)
>>>>            trace("test_query_value_ex: type set to: 0x%08x\n", type);
>>>>            ok(size == 0, "size should have been set to 0 instead of %d\n", size);
>>>>        }
>>>> +    todo_wine ok(GetLastError() == ERROR_SUCCESS, "expected ERROR_SUCCESS, got %d\n", GetLastError());
>>>>
>>>>        size = sizeof(buffer);
>>>> +    /* this last error check could be removed after setting it properly is fixed,
>>>> +       it's to prevent regressions from current state */
>>>> +    SetLastError(ERROR_SUCCESS);
>>>>        ret = RegQueryValueExA(HKEY_CLASSES_ROOT, "Nonexistent Value", NULL,&type, buffer,&size);
>>>>        ok(ret == ERROR_FILE_NOT_FOUND, "expected ERROR_FILE_NOT_FOUND, got %d\n", ret);
>>>>        ok(size == sizeof(buffer), "size shouldn't have been changed to %d\n", size);
>>>> +    ok(GetLastError() == ERROR_SUCCESS, "expected ERROR_SUCCESS, got %d\n", GetLastError());
>>>>
>>>>          
>>> This is confusing. You should set it to 0xdeadbeef before each call for
>>> the check to be meaningful.
>>>
>>>        
>> Ok, I'll remove this part. Reason was to point a case when check is
>> working with current code.
>>      
> I'm not saying that you should remove it, but that you should set last
> error to 0xdeadbeef to make it clear whether you are expecting it to be
> unchanged or set to 0. Right now your test could mean either.
>    
Previous check is about the same thing, I reset to deadbeef and check 
for 0 after call. Second one is what application does, I wanted it as a 
passing test to show that error isn't changed from 0. It's ok to remove it.