================
@@ -0,0 +1,99 @@
+// RUN: %clang_cc1 -triple i686-windows %s -fsyntax-only -Wno-unused-value \
+// RUN: -Wno-pointer-to-int-cast -Wmicrosoft -verify=ms-anonymous
-fms-anonymous-structs
+// RUN: %clang_cc1 -triple powerpc-ibm-aix %s -fsyntax-only -Wno-unused-value \
+// RUN: -Wno-pointer-to-int-cast -Wmicrosoft -verify=ms-anonymous
-fms-anonymous-structs
+// RUN: %clang_cc1 -triple i686-windows %s -fsyntax-only -Wno-unused-value \
+// RUN: -Wno-pointer-to-int-cast -Wmicrosoft -verify=ms-anonymous
-fms-extensions
+// RUN: %clang_cc1 -triple i686-windows %s -fsyntax-only -Wno-unused-value \
+// RUN: -Wno-pointer-to-int-cast -Wmicrosoft -verify=ms-anonymous
-fms-compatibility
+// RUN: %clang_cc1 -triple i686-windows %s -fsyntax-only -Wno-unused-value \
+// RUN: -Wno-pointer-to-int-cast -Wmicrosoft -verify=ms-anonymous-dis
+
+typedef struct notnested {
+ long bad1;
+ long bad2;
+} NOTNESTED;
+
+
+typedef struct nested1 {
+ long a;
+ struct notnested var1;
+ NOTNESTED var2;
----------------
hubert-reinterpretcast wrote:
The functional changes in the PR do not warrant testing in so much detail (that
members of non-anonymous structs cannot be referenced as members of the
containing struct). On the other hand, if we were interested in testing for
that, the more interesting case is if there were only one non-anonymous struct
member of each struct type.
I suggest reducing the test to only one simple MS anonymous struct case, one
simple MS anonymous union case, and one transitive case. In particular, omit
the cases where an error is expected even when the extension is enabled.
https://github.com/llvm/llvm-project/pull/176551
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
