JVM IR: prevent behavior change with operator dot calls on literals

#KT-42321 Fixed

compiler/testData/ir/irText/expressions/kt42321.kt

@@ -0,0 +1,62 @@
+// SKIP_KT_DUMP
+// This test exists only to check that we don't accidentally break the buggy behavior of the old JVM backend in JVM IR (KT-42321).
+// Feel free to remove it as soon as there's no language version where such code is allowed (KT-38895).
+
+abstract class C<L> {
+    abstract fun takeT(x: L)
+}
+
+fun testLongDotCall(c1: C<Long>) {
+    c1.takeT(1.plus(2))
+    c1.takeT(1.minus(2))
+    c1.takeT(1.times(2))
+    c1.takeT(1.div(2))
+    c1.takeT(1.rem(2))
+    c1.takeT(1.unaryPlus())
+    c1.takeT(1.unaryMinus())
+    c1.takeT(1.shl(2))
+    c1.takeT(1.shr(2))
+    c1.takeT(1.ushr(2))
+    c1.takeT(1.and(2))
+    c1.takeT(1.or(2))
+    c1.takeT(1.xor(2))
+    c1.takeT(1.inv())
+}
+
+fun testShortDotCall(c2: C<Short>) {
+    c2.takeT(1.plus(2))
+    c2.takeT(1.shr(2))
+    c2.takeT(1.inv())
+}
+
+fun testByteDotCall(c3: C<Byte>) {
+    c3.takeT(1.plus(2))
+    c3.takeT(1.shr(2))
+    c3.takeT(1.inv())
+}
+
+fun testLongOperatorInfixCall(c4: C<Long>) {
+    c4.takeT(1 + 2)
+    c4.takeT(1 - 2)
+    c4.takeT(1 * 2)
+    c4.takeT(1 / 2)
+    c4.takeT(1 % 2)
+    c4.takeT(+1)
+    c4.takeT(-1)
+    c4.takeT(1 shl 2)
+    c4.takeT(1 shr 2)
+    c4.takeT(1 ushr 2)
+    c4.takeT(1 and 2)
+    c4.takeT(1 or 2)
+    c4.takeT(1 xor 2)
+}
+
+fun testShortOperatorInfixCall(c5: C<Short>) {
+    c5.takeT(1 + 2)
+    c5.takeT(1 shr 2)
+}
+
+fun testByteOperatorInfixCall(c6: C<Byte>) {
+    c6.takeT(1 + 2)
+    c6.takeT(1 shr 2)
+}