[llvm] [JITLink][AArch32] Add dynamic lookup for relocation fixup infos (PR #71649)

[via llvm-commits]

Stefan =?utf-8?q?Gränitz?= <stefan.graenitz at gmail.com>,
Stefan =?utf-8?q?Gränitz?= <stefan.graenitz at gmail.com>,
Stefan =?utf-8?q?Gränitz?= <stefan.graenitz at gmail.com>,
Stefan =?utf-8?q?Gränitz?= <stefan.graenitz at gmail.com>
Message-ID:
In-Reply-To: <llvm.org/llvm/llvm-project/pull/71649 at github.com>


================
@@ -163,65 +164,107 @@ struct HalfWords {
   const uint16_t Lo; // Second halfword
 };
 
-/// Collection of named constants per fixup kind. It may contain but is not
-/// limited to the following entries:
+/// FixupInfo base class is required for dynamic lookups.
+struct FixupInfoBase {
+  static const FixupInfoBase *getDynFixupInfo(Edge::Kind K);
+  virtual ~FixupInfoBase() {}
+};
+
+/// FixupInfo checks for Arm edge kinds work on 32-bit words
+struct FixupInfoArm : public FixupInfoBase {
+  bool (*checkOpcode)(uint32_t Wd) = nullptr;
+};
+
+/// FixupInfo check for Thumb32 edge kinds work on a pair of 16-bit halfwords
+struct FixupInfoThumb : public FixupInfoBase {
+  bool (*checkOpcode)(uint16_t Hi, uint16_t Lo) = nullptr;
+};
+
+/// Collection of named constants per fixup kind
 ///
+/// Mandatory entries:
 ///   Opcode      - Values of the op-code bits in the instruction, with
 ///                 unaffected bits nulled
 ///   OpcodeMask  - Mask with all bits set that encode the op-code
+///
+/// Other common entries:
 ///   ImmMask     - Mask with all bits set that encode the immediate value
 ///   RegMask     - Mask with all bits set that encode the register
 ///
+/// Specializations can add further custom fields without restrictions.
+///
 template <EdgeKind_aarch32 Kind> struct FixupInfo {};
 
-template <> struct FixupInfo<Arm_Jump24> {
+namespace {
----------------
mikaelholmen wrote:

I've no idea. I don't see them with clang either.
But downstream we compile with both clang and gcc and I just happened to notice that a bunch of new warnings popped up now with gcc.
It's no big deal but if they'll be gone in some nearish future I'm happy since it makes it easier to spot possible real problems.
Thanks!

https://github.com/llvm/llvm-project/pull/71649