[PATCH] mfplat/tests: Print an error if the first call to get_event fails in test_source_resolver (Coverity)

[Alex Henrie]

The test will still segfault if it fails, but at least there will be an
indication as to why.

Signed-off-by: Alex Henrie 
---
 dlls/mfplat/tests/mfplat.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/dlls/mfplat/tests/mfplat.c b/dlls/mfplat/tests/mfplat.c
index 21eab6d1172..d9f30867f7f 100644
--- a/dlls/mfplat/tests/mfplat.c
+++ b/dlls/mfplat/tests/mfplat.c
@@ -686,6 +686,7 @@ static void test_source_resolver(void)
     GUID guid;
     float rate;
     UINT32 rotation;
+    BOOL ret;

if (!pMFCreateSourceResolver)
     {
@@ -889,12 +890,10 @@ todo_wine
     hr = IMFMediaSource_Start(mediasource, descriptor, &GUID_NULL, &var);
     ok(hr == S_OK, "Failed to start media source, hr %#x.\n", hr);

-    video_stream = NULL;
-    if (get_event((IMFMediaEventGenerator *)mediasource, MENewStream, &var))
-    {
-        ok(var.vt == VT_UNKNOWN, "Unexpected value type.\n");
-        video_stream = (IMFMediaStream *)var.punkVal;
-    }
+    ret = get_event((IMFMediaEventGenerator *)mediasource, MENewStream, &var);
+    ok(ret, "Failed to get event.\n");
+    ok(var.vt == VT_UNKNOWN, "Unexpected value type.\n");
+    video_stream = (IMFMediaStream *)var.punkVal;

sample_count = 10;

@@ -912,7 +911,6 @@ todo_wine
         LONGLONG duration, time;
         DWORD buffer_count;
         IMFSample *sample;
-        BOOL ret;

ret = get_event((IMFMediaEventGenerator *)video_stream, MEMediaSample, &var);
         ok(ret, "Sample %u not received.\n", i + 1);

[Nikolay Sivov]

On 12/27/21 02:11, Alex Henrie wrote:
> - video_stream = NULL;
> - if (get_event((IMFMediaEventGenerator *)mediasource, MENewStream, &var))
> - {
> -
>    ok(var.vt == VT_UNKNOWN, "Unexpected value type.\n");
> -
>    video_stream = (IMFMediaStream *)var.punkVal;
> - }
> - ret = get_event((IMFMediaEventGenerator *)mediasource, MENewStream, &var);
> - ok(ret, "Failed to get event.\n");
> - ok(var.vt == VT_UNKNOWN, "Unexpected value type.\n");
> - video_stream = (IMFMediaStream *)var.punkVal;

The point is that if it's not MENewStream, you won't necessarily have 
IMFMediaStream in event data.

[Alex Henrie]

On Sun, Dec 26, 2021 at 11:45 PM Nikolay Sivov  wrote:
> On 12/27/21 02:11, Alex Henrie wrote:
> > - video_stream = NULL;
> > - if (get_event((IMFMediaEventGenerator *)mediasource, MENewStream, &var))
> > - {
> > -
> >    ok(var.vt == VT_UNKNOWN, "Unexpected value type.\n");
> > -
> >    video_stream = (IMFMediaStream *)var.punkVal;
> > - }
> > - ret = get_event((IMFMediaEventGenerator *)mediasource, MENewStream, &var);
> > - ok(ret, "Failed to get event.\n");
> > - ok(var.vt == VT_UNKNOWN, "Unexpected value type.\n");
> > - video_stream = (IMFMediaStream *)var.punkVal;
>
> The point is that if it's not MENewStream, you won't necessarily have
> IMFMediaStream in event data.

If that ever happened, the next test would crash because video_stream
would be null (hence the Coverity warning). So, as they are currently
written, the tests assume that the first call to get_event will
succeed. Is there any version of Windows that we care about where the
first call to get_event fails?

-Alex