[PATCH] D53264: [ARM64] [Windows] Add unwind support to llvm-readobj

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a subscriber: compnerd.
efriedma added inline comments.


================
Comment at: tools/llvm-readobj/ARMWinEHPrinter.cpp:788
       }
-      assert(DI < array_lengthof(Ring) && "unhandled opcode");
+      if ((isAArch64 && (DI >= array_lengthof(Ring))) ||
+          (!isAArch64 && (DI >= array_lengthof(Ring)))) {
----------------
Ring64


================
Comment at: tools/llvm-readobj/ARMWinEHPrinter.cpp:790
+          (!isAArch64 && (DI >= array_lengthof(Ring)))) {
+        SW.startLine() << format("0x%02x             ;  Bad opcode!\n", Opcodes.data()[Offset]);
+        ++OI;
----------------
80 columns


================
Comment at: tools/llvm-readobj/ARMWinEHPrinter.cpp:791
+        SW.startLine() << format("0x%02x             ;  Bad opcode!\n", Opcodes.data()[Offset]);
+        ++OI;
+      }
----------------
`break;`


================
Comment at: tools/llvm-readobj/ARMWinEHPrinter.cpp:836
+    llvm::errs() << "llvm-readobj doesn't yet support COFF unwind info that "
+                    "requires Extended Epilog Count and Extended Code Words.\n";
+    return false;
----------------
I thought you implemented this...?  At least at first glance, ExceptionDataRecord tries to do the right thing.


================
Comment at: tools/llvm-readobj/ARMWinEHPrinter.cpp:840
+  if (Contents.size() - Offset - 4 -
+      (XData.E() ? 0 : XData.EpilogueCount() * 4) < ByteCodeLength)
+    report_fatal_error("Malformed unwind data");
----------------
I think the subtraction here can overflow.


Repository:
  rL LLVM

https://reviews.llvm.org/D53264