[InstCombine] Remove instructions before+after unreachable at same time

There is no need to first remove the instructions before and then
the ones after in two different worklist iterations. We don't need
to worry about change reporting here, as the functions do that
themselves.

This avoids the issue in #150338, but not really in a principled
way. It's possible that we will have to allow poison arguments
to lifetime.start/lifetime.end again if this turns out to be a
recurring problem.

Author: nikic

llvm/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp

@@ -1502,8 +1502,7 @@ Instruction *InstCombinerImpl::visitStoreInst(StoreInst &SI) {
   // This is a non-terminator unreachable marker. Don't remove it.
   if (isa<UndefValue>(Ptr)) {
     // Remove guaranteed-to-transfer instructions before the marker.
-    if (removeInstructionsBeforeUnreachable(SI))
-      return &SI;
+    removeInstructionsBeforeUnreachable(SI);
 
     // Remove all instructions after the marker and handle dead blocks this
     // implies.

llvm/test/Transforms/InstCombine/pr150338.ll

@@ -0,0 +1,16 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -passes=instcombine < %s | FileCheck %s
+
+; Make sure this does not crash.
+define void @test(ptr %arg) {
+; CHECK-LABEL: define void @test(
+; CHECK-SAME: ptr [[ARG:%.*]]) {
+; CHECK-NEXT:    store i1 true, ptr poison, align 1
+; CHECK-NEXT:    ret void
+;
+  %a = alloca i32
+  store ptr %a, ptr %arg
+  store i1 true, ptr poison
+  call void @llvm.lifetime.end.p0(i64 4, ptr %a)
+  ret void
+}