Don't assert int64_t alignment on x86.
Fix build.
Change-Id: I1b798bb3c5ab4954b46d54cda8f8c237ab4ae53e
diff --git a/runtime/atomic.h b/runtime/atomic.h
index 4ec1f4e..04daea8 100644
--- a/runtime/atomic.h
+++ b/runtime/atomic.h
@@ -561,8 +561,12 @@
COMPILE_ASSERT(alignof(AtomicInteger) == alignof(int32_t),
atomic_int_alignment_differs_from_that_of_underlying_type);
COMPILE_ASSERT(sizeof(Atomic<int64_t>) == sizeof(int64_t), weird_atomic_int64_size);
-COMPILE_ASSERT(alignof(Atomic<int64_t>) == alignof(int64_t),
- atomic_int64_alignment_differs_from_that_of_underlying_type);
+#if defined(__LP64__)
+ COMPILE_ASSERT(alignof(Atomic<int64_t>) == alignof(int64_t),
+ atomic_int64_alignment_differs_from_that_of_underlying_type);
+#endif
+// The above fails on x86-32.
+// This is OK, since we explicitly arrange for alignment of 8-byte fields.
#if !ART_HAVE_STDATOMIC
