fix issue 20042 - __vector CTFE crashes the compiler

src/dmd/dinterpret.d

@@ -4262,6 +4262,13 @@ public:
                 lowerbound = 0;
                 upperbound = 0;
             }
+            else if (VectorExp ve = e1.isVectorExp())
+            {
+                // ve is not handled but a proper error message is returned
+                // this is to prevent https://issues.dlang.org/show_bug.cgi?id=20042
+                lowerbound = 0;
+                upperbound = ve.dim;
+            }
             else
                 assert(0);
 

test/fail_compilation/ice20042.d

@@ -0,0 +1,29 @@
+/*
+DISABLED: freebsd32 linux32 osx32 win32
+TEST_OUTPUT:
+---
+fail_compilation/ice20042.d(18): Error: slice operation `cast(__vector(float[4]))nanF = [1.00000F, 2.00000F, 3.00000F, 4.00000F][0..4]` cannot be evaluated at compile time
+fail_compilation/ice20042.d(25):        called from here: `Vec4(cast(__vector(float[4]))[nanF, nanF, nanF, nanF]).this([1.00000F, 2.00000F, 3.00000F, 4.00000F])`
+---
+*/
+void write(T...)(T t){}
+
+struct Vec4
+{
+    __vector(float[4]) raw;
+
+    this(const(float[4]) value...) inout pure @safe nothrow @nogc
+    {
+        __vector(float[4]) raw;
+        raw[] = value[];
+        this.raw = raw;
+    }
+}
+
+void main()
+{
+    static immutable Vec4 v = Vec4(  1.0f, 2.0f, 3.0f, 4.0f );
+
+    static foreach(d; 0 .. 4)
+        write(v.raw[d], " ");
+}