stage2: LLVM backend: miscompilation fixes

 * work around a stage1 miscompilation leading to the wrong integer
   comparison predicate being emitted.
 * fix the bug of not annotating callsites with the calling convention
   of the callee, leading to undefined behavior.
 * add the `nobuiltin` attribute when building freestanding libc or
   compiler_rt libraries to prevent e.g. memcpy from being "optimized"
   into a call to itself.
 * compiler-rt: change a call to be comptime to make the generated LLVM
   IR simpler and easier to study.

I still can't enable the widening tests due to the compiler-rt compare
function being miscompiled in some not-yet-diagnosed way.

1 files changed, 1 insertions, 1 deletions

diff --git a/lib/std/special/compiler_rt/compareXf2.zig b/lib/std/special/compiler_rt/compareXf2.zig
index 951632b351..9f3750094e 100644
--- a/lib/std/special/compiler_rt/compareXf2.zig
+++ b/lib/std/special/compiler_rt/compareXf2.zig
@@ -32,7 +32,7 @@ pub inline fn cmp(comptime T: type, comptime RT: type, a: T, b: T) RT {
     const exponentBits = std.math.floatExponentBits(T);
     const signBit = (@as(rep_t, 1) << (significandBits + exponentBits));
     const absMask = signBit - 1;
-    const infT = std.math.inf(T);
+    const infT = comptime std.math.inf(T);
     const infRep = @bitCast(rep_t, infT);
 
     const aInt = @bitCast(srep_t, a);