[PATCH] windowscodecs: Fix an MSVC warning about pNumericRepresentation.

[Serge Gautherie]

Signed-off-by: Serge Gautherie <winehq-git_serge_180711(a)gautherie.fr>
---
"...\info.c(1461) : error C4133: 'function' : incompatible types - from 'WICPixelFormatNumericRepresentation *' to 'DWORD *'"

ReactOS-Bug: https://jira.reactos.org/browse/CORE-7538
---
 dlls/windowscodecs/info.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/dlls/windowscodecs/info.c b/dlls/windowscodecs/info.c
index e131107..21b9c5d 100644
--- a/dlls/windowscodecs/info.c
+++ b/dlls/windowscodecs/info.c
@@ -1450,11 +1450,16 @@ static HRESULT WINAPI PixelFormatInfo_SupportsTransparency(IWICPixelFormatInfo2
 static HRESULT WINAPI PixelFormatInfo_GetNumericRepresentation(IWICPixelFormatInfo2 *iface,
     WICPixelFormatNumericRepresentation *pNumericRepresentation)
 {
+    HRESULT hr;
+    DWORD numrep;
+
     PixelFormatInfo *This = impl_from_IWICPixelFormatInfo2(iface);
 
     TRACE("(%p,%p)\n", iface, pNumericRepresentation);
 
-    return ComponentInfo_GetDWORDValue(This->classkey, numericrepresentation_valuename, pNumericRepresentation);
+    hr = ComponentInfo_GetDWORDValue(This->classkey, numericrepresentation_valuename, &numrep);
+    *pNumericRepresentation = numrep;
+    return hr;
 }
 
 static const IWICPixelFormatInfo2Vtbl PixelFormatInfo_Vtbl = {
-- 
2.10.0.windows.1

[Vincent Povirk]

This fails the tests because it eliminates the NULL check for
pNumericRepresentation which was handled in
ComponentInfo_GetDWORDValue. It should probably also not be assigned
if ComponentInfo_GetDWORDValue fails.

[Vincent Povirk]

BTW, WICPixelFormatNumericRepresentation is DWORD-sized (and if not we
have a binary compatibility problem), so it should be OK to cast the
pointer.

[Serge Gautherie]

On 26/07/2018 22:32, Vincent Povirk wrote:
> BTW, WICPixelFormatNumericRepresentation is DWORD-sized (and if not we
> have a binary compatibility problem), so it should be OK to cast the
> pointer.

According to WikiPedia, "C exposes the integer representation of 
enumeration values ...". As in enum type is somewhat obscure.
https://en.wikipedia.org/wiki/Enumerated_type#C

To be on the extra-safe/explicit side, I would use an intermediate DWORD.

--

[Vincent Povirk]

If the enum is for some reason not DWORD-sized, an intermediate
variable would get us the wrong behavior for binary compatibility.
We'd be assigning a different sized value to the caller's DWORD-sized
variable. Of course, casting would break if Wine or winelib calls the
function internally.

It seems to me that neither approach is really safer, but casting
allows for better code sharing.

[Serge Gautherie]

On 26/07/2018 22:30, Vincent Povirk wrote:
> This fails the tests because it eliminates the NULL check for
> pNumericRepresentation which was handled in
> ComponentInfo_GetDWORDValue. It should probably also not be assigned
> if ComponentInfo_GetDWORDValue fails.

I concur. Sorry for not having paid enough attention.

--