Merging r243384:
------------------------------------------------------------------------
r243384 | slthakur | 2015-07-28 06:51:51 +0100 (Tue, 28 Jul 2015) | 9 lines

[UBSan][MIPS] Fix cast-overflow tests for mips big endian

This fixes the bug https://llvm.org/bugs/show_bug.cgi?id=24152
The float value resides in the first 4 bytes of ValueHandle for both mips and mipsel.

Reviewers: dsanders, samsonov
Subscibers: rsmith, hans, mohit.bhakkad, jaydeep, llvm-commits
Differential: http://reviews.llvm.org/D11448

------------------------------------------------------------------------


git-svn-id: https://llvm.org/svn/llvm-project/compiler-rt/branches/release_37@245110 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/ubsan/ubsan_value.cc b/lib/ubsan/ubsan_value.cc
index 215c195..8e298d4 100644
--- a/lib/ubsan/ubsan_value.cc
+++ b/lib/ubsan/ubsan_value.cc
@@ -83,10 +83,11 @@
 #endif
       case 32: {
         float Value;
-#if defined(__BIG_ENDIAN__)
-       // For big endian the float value is in the second 4 bytes
-       //  instead of the first 4 bytes.
-       internal_memcpy(&Value, ((const char*)&Val)+4, 4);
+#if defined(__BIG_ENDIAN__) && !defined(__mips__)
+       // For big endian the float value is in the highest addressed 4 bytes
+       // (the least significant bits) but we must also consider that we might
+       // have a 32-bit type.
+       internal_memcpy(&Value, ((const char*)&Val + 1) - 4, 4);
 #else 
        internal_memcpy(&Value, &Val, 4);
 #endif