[resend PATCH] ntdll: Tolerate null handle in DeregisterWait

[Keno Fischer]

Thanks for committing this. I was going to address the review
comments, but didn't get to that before this was committed. Would you
like me to submit a separate patch to address the review comments?

On Fri, Aug 19, 2016 at 5:57 PM, Bruno Jesus <00cpxxx at gmail.com> wrote:
> On Fri, Aug 19, 2016 at 6:24 PM, Keno Fischer <keno at juliacomputing.com> wrote:
>
> Hi, is the subject supposed to be Tolerate NULL handle in UnregisterWait?
>
>> diff --git a/dlls/kernel32/tests/thread.c b/dlls/kernel32/tests/thread.c
>> index a9a04df..44c6278 100644
>  > --- a/dlls/kernel32/tests/thread.c
>> +++ b/dlls/kernel32/tests/thread.c
>> @@ -1245,6 +1245,11 @@ static void test_RegisterWaitForSingleObject(void)
>>
>>      ret = pUnregisterWait(wait_handle);
>>      ok(ret, "UnregisterWait failed with error %d\n", GetLastError());
>> +
>> +    /* Test UnregisterWait null argument case */
>> +    ret = pUnregisterWait(0);
>> +    ok(!ret && GetLastError() == ERROR_INVALID_HANDLE,
>> +        "UnregisterWait failed with unexpected error %d\n", GetLastError());
>>  }
>
> Please test ret and GetLastError separately as the rest of the code
> does most times.
>
> If you see other tests that test GetLastError you will notice that
> they call SetLastError before calling the tested function to ensure
> the error is not from a previous test. So add a
> SetLastError(0xdeadbeef) before the call to pUnregisterWait to ensure
> the ERROR_INVALID_HANDLE returned is from UnregisterWait.
>
> Also for consistency why not pUnregisterWait(NULL) instead of (0)?