Add a test case to showcase a difference in #pragma pack handling

compared to gcc. This is worrisome, but I believe we are doing the
"correct" thing, and if I recall correctly I previously verified this
versus MSVC.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@71723 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/Sema/pragma-pack-3.c b/test/Sema/pragma-pack-3.c
new file mode 100644 (file)
index 0000000..a2d665e
--- /dev/null
+++ b/test/Sema/pragma-pack-3.c
@@ -0,0 +1,34 @@
+// RUN: clang-cc -triple i686-apple-darwin9 %s -fsyntax-only -verify
+
+// Stack: [], Alignment: 8
+
+#pragma pack(push, 1)
+// Stack: [8], Alignment: 1
+
+#pragma pack(push, 4)
+// Stack: [8, 1], Alignment: 4
+
+// Note that this differs from gcc; pack() in gcc appears to pop the
+// top stack entry and resets the current alignment. This is both
+// inconsistent with MSVC, and the gcc documentation. In other cases,
+// for example changing this to pack(8), I don't even understand what gcc
+// is doing.
+
+#pragma pack()
+// Stack: [8, 1], Alignment: 8
+
+#pragma pack(pop)
+// Stack: [8], Alignment: 1
+struct s0 {
+  char f0;
+  short f1;
+};
+int a[sizeof(struct s0) == 3 ? 1 : -1];
+
+#pragma pack(pop)
+// Stack: [], Alignment: 8
+struct s1 {
+  char f0;
+  short f1;
+};
+int b[sizeof(struct s1) == 4 ? 1 : -1];