[MIPS] Add remaining parts for building tests for MIPS

tests/exception_test/exception_test.c

Index: tests/exception_test/exception_test.c
diff --git a/tests/exception_test/exception_test.c b/tests/exception_test/exception_test.c
index 7c56582204fd857ce0c65a515db46785c4d8ce01..310593ad2e198ff1fd2ddbcb52178687ddbf7aef 100644
--- a/tests/exception_test/exception_test.c
+++ b/tests/exception_test/exception_test.c
@@ -63,7 +63,11 @@ struct AlignedType {
  */
 __attribute__((noinline))
 void check_pointer_is_aligned(void *pointer) {
+#if defined(__mips__)
+  assert((uintptr_t) pointer % 8 == 0);

[Mark Seaborn, 2013/02/13 22:14:09]

But we declared it with __attribute__((aligned(16))), so no conditional should
be needed for this to pass

[petarj, 2013/03/05 17:50:04]

attribute aligned does not work for variables on stack on MIPS.
Speaking of which, what is the meaning of check_stack_is_aligned
if you are putting a 16-byte aligned struct on stack and then
checking if it is aligned? It looks more like you are checking that
attribute aligned works correctly rather than checking that stack
pointer is aligned.

[Mark Seaborn, 2013/03/14 15:48:00]

Answering your earlier question...

On 2013/03/05 17:50:04, petarj wrote:
Do you mean it does not actually align the address of the variable?  That would
be a bug, right?  Do you mean that the "align" attribute of LLVM's "alloca"
instruction doesn't work on MIPS?

Or do you just mean the backend does not make any assumptions about the
alignment of the stack?
No, we are checking that the stack pointer is aligned, but the check only works
if the compiler makes use of the ABI's stack alignment and does not force a
realignment of the stack pointer.

So this check isn't very reliable, and you are welcome to remove
check_stack_is_aligned().  This check is made obsolete by the later check,
assert(frame_base % STACK_ALIGNMENT == 0).

+#else
   assert((uintptr_t) pointer % 16 == 0);
+#endif
 }
 
 void check_stack_is_aligned(void) {