Fix chrome/arm build warnings.

third_party/qcms/src/transform.c

Index: third_party/qcms/src/transform.c
diff --git a/third_party/qcms/src/transform.c b/third_party/qcms/src/transform.c
index 44ce9b5631f047866ae69c21e8c8ec8cdeb6ebaa..fdab82cbfe0da31c0f7e2240183711ee48c07cb8 100644
--- a/third_party/qcms/src/transform.c
+++ b/third_party/qcms/src/transform.c
@@ -1323,19 +1323,9 @@ qcms_transform* qcms_transform_create(
 	return transform;
 }
 
-#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__)
-#if !defined(__has_attribute)
-/* Assume that the compiler supports the provided attribute. */
-#define __has_attribute(x) 1
-#endif
-#if __has_attribute(__force_align_arg_pointer__)
+#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__) && !defined(__arm__)

[Nico, 2012/11/13 22:53:49]

Instead of removing all the __has_attribute stuff, I'd keep that and be like

# if __has__atribute(__force_align…) || what you have.

I'd also do the two changes in two CLs, as they aren't related.

[Sam Clegg, 2012/11/13 23:24:46]

You mean have __has_attribute() default to 0 when
building with non-clang?

It seems to me that the __has_attribute() stuff is clang
only.  What do we gain by including it rather the simple
single line patch?

[Nico, 2012/11/13 23:29:01]

It has the advantage that it asks the compiler if it knows about the attribute
(if the compiler supports that, which currently only clang does). Feels a bit
more future-proof.

 /* we need this to avoid crashes when gcc assumes the stack is 128bit aligned */
 __attribute__((__force_align_arg_pointer__))
-#else
-/* __force_align_arg_pointer__ is an x86-only attribute, and clang warns on used
- * attributes. Don't use this on ARM.
- */
-#endif
 #endif
 void qcms_transform_data(qcms_transform *transform, void *src, void *dest, size_t length)
 {