[PATCH 0/3] MR6502: include: Fix MIDL_INTERFACE definition.

[Rémi Bernon (＠rbernon)]

Clang in MSVC mode supports this and __uuidof fails / complains about missing virtual destructor otherwise.

[PATCH 1/3] include: Use winnt.h DECLSPEC_UUID definition in rpcndr.h.

[Rémi Bernon]

From: Rémi Bernon

---
 include/rpcndr.h | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/include/rpcndr.h b/include/rpcndr.h
index 93ebb0a4aa7..9fcdf706709 100644
--- a/include/rpcndr.h
+++ b/include/rpcndr.h
@@ -112,7 +112,14 @@ typedef void (__RPC_USER *NDR_RUNDOWN)(void *context);
 typedef void (__RPC_USER *NDR_NOTIFY_ROUTINE)(void);
 typedef void (__RPC_USER *NDR_NOTIFY2_ROUTINE)(boolean flag);

-#define DECLSPEC_UUID(x)
+#ifndef DECLSPEC_UUID
+# if defined(_MSC_VER) && (_MSC_VER >= 1100) && defined (__cplusplus)
+#  define DECLSPEC_UUID(x) __declspec(uuid(x))
+# else
+#  define DECLSPEC_UUID(x)
+# endif
+#endif
+
 #define MIDL_INTERFACE(x)   struct

struct _MIDL_STUB_MESSAGE;

[PATCH 2/3] include: Use winnt.h DECLSPEC_NOVTABLE definition in rpcndr.h.

[Rémi Bernon]

From: Rémi Bernon

---
 include/rpcndr.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/include/rpcndr.h b/include/rpcndr.h
index 9fcdf706709..de6a0663650 100644
--- a/include/rpcndr.h
+++ b/include/rpcndr.h
@@ -112,6 +112,14 @@ typedef void (__RPC_USER *NDR_RUNDOWN)(void *context);
 typedef void (__RPC_USER *NDR_NOTIFY_ROUTINE)(void);
 typedef void (__RPC_USER *NDR_NOTIFY2_ROUTINE)(boolean flag);

+#ifndef DECLSPEC_NOVTABLE
+# if defined(_MSC_VER) && (_MSC_VER >= 1100) && defined(__cplusplus)
+#  define DECLSPEC_NOVTABLE __declspec(novtable)
+# else
+#  define DECLSPEC_NOVTABLE
+# endif
+#endif
+
 #ifndef DECLSPEC_UUID
 # if defined(_MSC_VER) && (_MSC_VER >= 1100) && defined (__cplusplus)
 #  define DECLSPEC_UUID(x) __declspec(uuid(x))

[PATCH 3/3] include: Add DECLSPEC_UUID/DECLSPEC_NOVTABLE to MIDL_INTERFACE.

[Rémi Bernon]

From: Rémi Bernon

---
 include/rpcndr.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/rpcndr.h b/include/rpcndr.h
index de6a0663650..4d140492558 100644
--- a/include/rpcndr.h
+++ b/include/rpcndr.h
@@ -128,7 +128,7 @@ typedef void (__RPC_USER *NDR_NOTIFY2_ROUTINE)(boolean flag);
 # endif
 #endif

-#define MIDL_INTERFACE(x)   struct
+#define MIDL_INTERFACE(x)   struct DECLSPEC_UUID(x) DECLSPEC_NOVTABLE

struct _MIDL_STUB_MESSAGE;
 struct _MIDL_STUB_DESC;

[Jacek Caban (＠jacek)]

Jacek Caban (@jacek) commented about include/rpcndr.h:
> typedef void (__RPC_USER *NDR_NOTIFY_ROUTINE)(void);
>  typedef void (__RPC_USER *NDR_NOTIFY2_ROUTINE)(boolean flag);
>
> -#define DECLSPEC_UUID(x)
> +#ifndef DECLSPEC_UUID
> +# if defined(_MSC_VER) && (_MSC_VER >= 1100) && defined (__cplusplus)

It's also supported on mingw targets when `-fms-extensions` is used. I think that using `__has_declspec_attribute` would be more portable and precise. Unfortunately, it's not supported on the actual MSVC, but maybe it's worth having something like:
```
#ifndef __has_declspec_attribute
# #ifdef _MSC_VER
#  define __has_declspec_attribute(x) 1
# else
#  define __has_declspec_attribute(x) 0
# endif
#endif
```

[Rémi Bernon (＠rbernon)]

On Mon Sep 16 10:36:18 2024 +0000, Jacek Caban wrote:
> It's also supported on mingw targets when `-fms-extensions` is used. I
> think that using `__has_declspec_attribute` would be more portable and
> precise. Unfortunately, it's not supported on the actual MSVC, but maybe
> it's worth having something like:
> #ifndef __has_declspec_attribute
> # #ifdef _MSC_VER
> #  define __has_declspec_attribute(x) 1
> # else
> #  define __has_declspec_attribute(x) 0
> # endif
> #endif

Several of the DECLSPEC definitions also have additional conditions like _MSC_VER >= x, how would we handle these? Something like:

```
# if __has_declspec_attribute(noreturn) && (!defined(_MSC_VER) || (_MSC_VER >= 1200)) && !defined(MIDL_PASS)
```

Only looks worse.

[Jacek Caban (＠jacek)]

On Mon Sep 16 11:02:46 2024 +0000, Rémi Bernon wrote:
> Several of the DECLSPEC definitions also have additional conditions like
> _MSC_VER >= x, how would we handle these? Something like:
> # if __has_declspec_attribute(noreturn) && (!defined(_MSC_VER) ||
> (_MSC_VER >= 1200)) && !defined(MIDL_PASS)
>
> Only looks worse.

In this case, using `__attribute__((noreturn))` should be just fine, so I don't think we need to change that at all.

[Rémi Bernon (＠rbernon)]

On Mon Sep 16 11:30:30 2024 +0000, Jacek Caban wrote:
> In this case, using `__attribute__((noreturn))` should be just fine, so
> I don't think we need to change that at all.

Well this one maybe, but there are others. DECLSPEC_UUID requires _MSC_VER >= 1100, should we drop that condition altogether? What about later, for instance (taking one without alternatives) DECLSPEC_ADDRSAFE which has a different version requirement?

[Alexandre Julliard (＠julliard)]

On Mon Sep 16 11:54:44 2024 +0000, Rémi Bernon wrote:
> Well this one maybe, but there are others. DECLSPEC_UUID requires
> _MSC_VER >= 1100, should we drop that condition altogether? What about
> later, for instance (taking one without alternatives) DECLSPEC_ADDRSAFE
> which has a different version requirement?

Yes, these can be dropped. We don't need to remain compatible with MSVC from 25 years ago.