[llvm] [X86] Respect blockaddress offsets when performing X86 LEA fixups (PR #71641)

[Maurice Heumann via llvm-commits]

================
@@ -0,0 +1,30 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3
+# RUN: llc -mtriple=i386-unknown-linux-gnu -mattr=slow-3ops-lea -run-pass x86-fixup-LEAs -o -  %s | FileCheck %s
+
+--- |
+  define i32 @square(i32 %0) local_unnamed_addr {
----------------
momo5502 wrote:

Ah yeah, I understand, but no, unfortunately, I do not. I have a huge cpp file where I can reproduce the issue, but it is way too big and I'm not allowed to disclose it :/

I was not able to reproduce the issue with anything else than this cpp file. Running LLC and an IR file by itself, did not work. Somewhere in the lowering, the IR code would get transformed differently so that the bug did not occur. 

This is why I crafted this MIR test case :D

https://github.com/llvm/llvm-project/pull/71641