[llvm] include Compiler.h in a few headers where it was missed

[andrurogerz]

Add missing #include "llvm/Support/Compiler.h" in a few LLVM headers that use the LLVM_ABI macro.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Andrew Rogers (andrurogerz)

Changes

Add missing #include "llvm/Support/Compiler.h" in a few LLVM headers that use the LLVM_ABI macro.

---

Full diff: https://github.com/llvm/llvm-project/pull/144464.diff

3 Files Affected:

• (modified) llvm/include/llvm/Option/OptSpecifier.h (+2)
• (modified) llvm/include/llvm/Transforms/IPO.h (+2)
• (modified) llvm/include/llvm/Transforms/Utils/PromoteMemToReg.h (+2)

diff --git a/llvm/include/llvm/Option/OptSpecifier.h b/llvm/include/llvm/Option/OptSpecifier.h
index dc6acae7fc002..cb87fbd17ec1c 100644
--- a/llvm/include/llvm/Option/OptSpecifier.h
+++ b/llvm/include/llvm/Option/OptSpecifier.h
@@ -9,6 +9,8 @@
 #ifndef LLVM_OPTION_OPTSPECIFIER_H
 #define LLVM_OPTION_OPTSPECIFIER_H
 
+#include "llvm/Support/Compiler.h"
+
 namespace llvm {
 namespace opt {
 
diff --git a/llvm/include/llvm/Transforms/IPO.h b/llvm/include/llvm/Transforms/IPO.h
index 56b30968ffd77..7523ae66429ac 100644
--- a/llvm/include/llvm/Transforms/IPO.h
+++ b/llvm/include/llvm/Transforms/IPO.h
@@ -14,6 +14,8 @@
 #ifndef LLVM_TRANSFORMS_IPO_H
 #define LLVM_TRANSFORMS_IPO_H
 
+#include "llvm/Support/Compiler.h"
+
 namespace llvm {
 
 class ModulePass;
diff --git a/llvm/include/llvm/Transforms/Utils/PromoteMemToReg.h b/llvm/include/llvm/Transforms/Utils/PromoteMemToReg.h
index ab0bd3a5a9962..f20ae1809aa56 100644
--- a/llvm/include/llvm/Transforms/Utils/PromoteMemToReg.h
+++ b/llvm/include/llvm/Transforms/Utils/PromoteMemToReg.h
@@ -14,6 +14,8 @@
 #ifndef LLVM_TRANSFORMS_UTILS_PROMOTEMEMTOREG_H
 #define LLVM_TRANSFORMS_UTILS_PROMOTEMEMTOREG_H
 
+#include "llvm/Support/Compiler.h"
+
 namespace llvm {
 
 template <typename T> class ArrayRef;

[fahadnayyar]

I'm curious if there's a way to check if this change would unblock the GreenDragon bot failure. I'm concerned that it might introduce other regression issues in the GreenDragon bots. I wonder if we could try some PR testing to see if the bots don't regress further after this.