More complete support for opaque pointers (required for LLVM 15+)

.github/workflows/llvm-quick-fuzz.yml

@@ -23,7 +23,9 @@ jobs:
         os: [ubuntu-22.04]
         # See doc/developing.md
         ghc: ["8.10.7"]
-        llvm: [ "https://github.com/llvm/llvm-project/releases/download/llvmorg-14.0.0/clang+llvm-14.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz"
+        llvm: [ "https://github.com/llvm/llvm-project/releases/download/llvmorg-16.0.2/clang+llvm-16.0.2-x86_64-linux-gnu-ubuntu-22.04.tar.xz"
+              , "https://github.com/llvm/llvm-project/releases/download/llvmorg-15.0.6/clang+llvm-15.0.6-x86_64-linux-gnu-ubuntu-18.04.tar.xz"
+              , "https://github.com/llvm/llvm-project/releases/download/llvmorg-14.0.0/clang+llvm-14.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz"
               , "https://github.com/llvm/llvm-project/releases/download/llvmorg-13.0.0/clang+llvm-13.0.0-x86_64-linux-gnu-ubuntu-20.04.tar.xz"
               , "https://github.com/llvm/llvm-project/releases/download/llvmorg-12.0.0/clang+llvm-12.0.0-x86_64-linux-gnu-ubuntu-20.04.tar.xz"
               , "https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/clang+llvm-11.0.0-x86_64-linux-gnu-ubuntu-20.04.tar.xz"

README.md

@@ -32,8 +32,8 @@ compilers.
 |           | v12.0   | ✓                             |              |                      |
 |           | v13.0   | ✓                             |              | See [issues][llvm13] |
 |           | v14.0   | ✓                             |              | See [issues][llvm14] |
-|           | v15.0   |                               |              | See [issues][llvm15] |
-|           | v16.0   |                               |              | See [issues][llvm16] |
+|           | v15.0   | ✓                             |              | See [issues][llvm15] |
+|           | v16.0   | ✓                             |              | See [issues][llvm16] |
 | `clang++` | v3.4    |                               |              |                      |
 |           | v3.5    |                               |              |                      |
 |           | v3.6    |                               |              |                      |

disasm-test/Main.hs

@@ -215,6 +215,7 @@ cube = TS.mkCUBE
                                            , "pre-llvm12"
                                            , "pre-llvm13"
                                            , "pre-llvm14"
+                                           , "pre-llvm15"
                                            ])
                      ]
     -- Somewhat unusually for tasty-sugar, we make the expectedSuffix the same
@@ -366,11 +367,12 @@ processBitCode _keep (Roundtrip roundtrip) pfx file = do
   case e of
     Left err -> X.throwIO (ParseError err)
     Right m  -> do
-      parsed <- printToTempFile "ll" (show (ppLLVM (ppModule m)))
+      let m' = AST.fixupOpaquePtrs m
+      parsed <- printToTempFile "ll" (show (ppLLVM (ppModule m')))
       -- stripComments _keep parsed
       if roundtrip
       then do
-        tmp2 <- printToTempFile "ast" (ppShow (normalizeModule m))
+        tmp2 <- printToTempFile "ast" (ppShow (normalizeModule m'))
         return (parsed, Just tmp2)
       else return (parsed, Nothing)
 

disasm-test/tests/callbr.c

@@ -0,0 +1,18 @@
+// Adapted from the Clang test suite:
+// (https://github.com/llvm/llvm-project/blob/32103608fc073700f04238872d8b22655ddec3fb/clang/test/CodeGen/asm-goto.c#L5-L20),
+// which is licensed under the Apache License v2.0
+
+
+int main(void) {
+  int cond = 0;
+  asm volatile goto("testl %0, %0; jne %l1;"
+                    : /* No outputs */
+                    : "r"(cond)
+                    : /* No clobbers */
+                    : label_true, loop);
+  return 0;
+loop:
+  return 0;
+label_true:
+  return 1;
+}

disasm-test/tests/callbr.ll

@@ -1,40 +1,47 @@
-; ModuleID = 'test.c'
-source_filename = "test.c"
+; ModuleID = 'callbr.c'
+source_filename = "callbr.c"
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
-target triple = "x86_64-pc-linux-gnu"
+target triple = "x86_64-unknown-linux-gnu"
 
 ; Function Attrs: noinline nounwind optnone uwtable
-define dso_local i32 @test1(i32 %0) #0 {
+define dso_local i32 @main() #0 {
+  %1 = alloca i32, align 4
   %2 = alloca i32, align 4
-  %3 = alloca i32, align 4
-  store i32 %0, i32* %3, align 4
-  %4 = load i32, i32* %3, align 4
-  callbr void asm sideeffect "testl $0, $0; jne ${1:l};", "r,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %4, i8* blockaddress(@test1, %7), i8* blockaddress(@test1, %6)) #1
-          to label %5 [label %7, label %6], !srcloc !2
-
-5:                                                ; preds = %1
-  store i32 0, i32* %2, align 4
-  br label %8
-
-6:                                                ; preds = %1
-  store i32 0, i32* %2, align 4
-  br label %8
-
-7:                                                ; preds = %1
-  store i32 1, i32* %2, align 4
-  br label %8
-
-8:                                                ; preds = %7, %6, %5
-  %9 = load i32, i32* %2, align 4
-  ret i32 %9
+  store i32 0, ptr %1, align 4
+  store i32 0, ptr %2, align 4
+  %3 = load i32, ptr %2, align 4
+  callbr void asm sideeffect "testl $0, $0; jne ${1:l};", "r,!i,!i,~{dirflag},~{fpsr},~{flags}"(i32 %3) #1
+          to label %4 [label %6, label %5], !srcloc !7
+
+4:                                                ; preds = %0
+  store i32 0, ptr %1, align 4
+  br label %7
+
+5:                                                ; preds = %0
+  store i32 0, ptr %1, align 4
+  br label %7
+
+6:                                                ; preds = %0
+  store i32 1, ptr %1, align 4
+  br label %7
+
+7:                                                ; preds = %6, %5, %4
+  %8 = load i32, ptr %1, align 4
+  ret i32 %8
 }
 
-attributes #0 = { noinline nounwind optnone uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #0 = { noinline nounwind optnone uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 attributes #1 = { nounwind }
 
-!llvm.module.flags = !{!0}
-!llvm.ident = !{!1}
+!llvm.module.flags = !{!0, !1, !2, !3, !4}
+!llvm.ident = !{!5}
+!llvm.commandline = !{!6}
 
 !0 = !{i32 1, !"wchar_size", i32 4}
-!1 = !{!"clang version 10.0.0-4ubuntu1 "}
-!2 = !{i32 265}
+!1 = !{i32 7, !"PIC Level", i32 2}
+!2 = !{i32 7, !"PIE Level", i32 2}
+!3 = !{i32 7, !"uwtable", i32 2}
+!4 = !{i32 7, !"frame-pointer", i32 2}
+!5 = !{!"clang version 15.0.6"}
+!6 = !{!"/home/ryanglscott/Software/clang+llvm-15.0.6/bin/clang-15 -S -emit-llvm -frecord-command-line callbr.c -o callbr.ll"}
+!7 = !{i64 272}

disasm-test/tests/callbr.pre-llvm15.ll

@@ -0,0 +1,40 @@
+; ModuleID = 'test.c'
+source_filename = "test.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-pc-linux-gnu"
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i32 @test1(i32 %0) #0 {
+  %2 = alloca i32, align 4
+  %3 = alloca i32, align 4
+  store i32 %0, i32* %3, align 4
+  %4 = load i32, i32* %3, align 4
+  callbr void asm sideeffect "testl $0, $0; jne ${1:l};", "r,X,X,~{dirflag},~{fpsr},~{flags}"(i32 %4, i8* blockaddress(@test1, %7), i8* blockaddress(@test1, %6)) #1
+          to label %5 [label %7, label %6], !srcloc !2
+
+5:                                                ; preds = %1
+  store i32 0, i32* %2, align 4
+  br label %8
+
+6:                                                ; preds = %1
+  store i32 0, i32* %2, align 4
+  br label %8
+
+7:                                                ; preds = %1
+  store i32 1, i32* %2, align 4
+  br label %8
+
+8:                                                ; preds = %7, %6, %5
+  %9 = load i32, i32* %2, align 4
+  ret i32 %9
+}
+
+attributes #0 = { noinline nounwind optnone uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind }
+
+!llvm.module.flags = !{!0}
+!llvm.ident = !{!1}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{!"clang version 10.0.0-4ubuntu1 "}
+!2 = !{i32 265}

disasm-test/tests/opaque-atomicrmw.ll

@@ -0,0 +1,4 @@
+define void @atomicrmw(ptr %a, i32 %i) {
+    %b = atomicrmw add ptr %a, i32 %i acquire
+    ret void
+}

disasm-test/tests/opaque-atomicrmw.pre-llvm15.ll

@@ -0,0 +1,4 @@
+SKIP_TEST
+
+This test case requires the use of opaque pointers, which are most easily
+usable with LLVM 15 or later.

disasm-test/tests/opaque-call.ll

@@ -0,0 +1,11 @@
+define void @f(i32 %x) {
+    ret void
+}
+
+define void @g() {
+    %p = alloca ptr
+    store ptr @f, ptr %p
+    %f = load ptr, ptr %p
+    call void (i32) %f(i32 42)
+    ret void
+}

disasm-test/tests/opaque-call.pre-llvm15.ll

@@ -0,0 +1,4 @@
+SKIP_TEST
+
+This test case requires the use of opaque pointers, which are most easily
+usable with LLVM 15 or later.

disasm-test/tests/opaque-constant-getelementptr.ll

@@ -0,0 +1,9 @@
+%struct.RT = type { i8, [10 x [20 x i32]], i8 }
+%struct.ST = type { i32, double, %struct.RT }
+
+@.s = private constant %struct.ST zeroinitializer
+
+define ptr @foo() {
+entry:
+  ret ptr getelementptr inbounds (%struct.ST, ptr @.s, i64 1, i32 2, i32 1, i64 5, i64 13)
+}

disasm-test/tests/opaque-constant-getelementptr.pre-llvm15.ll

@@ -0,0 +1,4 @@
+SKIP_TEST
+
+This test case requires the use of opaque pointers, which are most easily
+usable with LLVM 15 or later.

disasm-test/tests/opaque-getelementptr.ll

@@ -0,0 +1,8 @@
+%struct.RT = type { i8, [10 x [20 x i32]], i8 }
+%struct.ST = type { i32, double, %struct.RT }
+
+define ptr @foo(ptr %s) {
+entry:
+  %arrayidx = getelementptr inbounds %struct.ST, ptr %s, i64 1, i32 2, i32 1, i64 5, i64 13
+  ret ptr %arrayidx
+}

disasm-test/tests/opaque-getelementptr.pre-llvm15.ll

@@ -0,0 +1,4 @@
+SKIP_TEST
+
+This test case requires the use of opaque pointers, which are most easily
+usable with LLVM 15 or later.

llvm-pretty-bc-parser.cabal

@@ -135,7 +135,7 @@ Test-suite disasm-test
                        tasty-hunit,
                        tasty-sugar >= 2.2 && < 2.3,
                        text,
-                       versions < 6,
+                       versions < 7,
                        llvm-pretty,
                        llvm-pretty-bc-parser
 

src/Data/LLVM/BitCode/Assert.hs

@@ -22,7 +22,6 @@ module Data.LLVM.BitCode.Assert
   -- ** Types
   , elimPtrTo
   , elimPtrTo_
-  , ptrTo
   ) where
 
 import           Control.Monad (MonadPlus, mplus)
@@ -32,7 +31,7 @@ import           Control.Monad.Fail (MonadFail)
 #endif
 import           Data.LLVM.BitCode.Record (Record)
 import qualified Data.LLVM.BitCode.Record as Record
-import           Text.LLVM.AST (Type', Typed, Ident)
+import           Text.LLVM.AST (Type', Ident)
 import qualified Text.LLVM.AST as AST
 
 supportedCompilerMessage :: [String]
@@ -80,43 +79,37 @@ recordSizeIn record ns =
 ----------------------------------------------------------------
 -- ** Types
 
--- | Assert that this thing is a pointer, get the underlying type
+-- | Assert that this thing is a @'PtrTo' ty@ and return the underlying @ty@.
+--
+-- Think carefully before using this function, as it will not work as you would
+-- expect when the type is an opaque pointer.
+-- See @Note [Pointers and pointee types]@.
 elimPtrTo :: (MonadFail m, MonadPlus m) => String -> Type' Ident -> m (Type' Ident)
 elimPtrTo msg ptrTy = AST.elimPtrTo ptrTy `mplus`
                         (fail $ unlines [ msg
                                         , "Expected pointer type, found:"
                                         , show ptrTy
                                         ])
 
--- | Assert that this thing is a pointer
+-- | Assert that this thing is a 'PtrTo' type.
+--
+-- Think carefully before using this function, as it will not work as you would
+-- expect when the type is an opaque pointer.
+-- See @Note [Pointers and pointee types]@.
 elimPtrTo_ :: (MonadFail m, MonadPlus m) => String -> Type' Ident -> m ()
 elimPtrTo_ msg ptrTy = elimPtrTo msg ptrTy >> pure ()
 
--- | Assert that the first thing is a pointer to something of the type of the
--- second thing, e.g. in a load/store instruction.
---
--- See: https://github.com/llvm-mirror/llvm/blob/release_60/lib/Bitcode/Reader/BitcodeReader.cpp#L3328
-ptrTo :: (MonadFail m, Show a, Show b)
-      => String
-      -> Typed a -- ^ The pointer
-      -> Typed b -- ^ The value
-      -> m ()
-ptrTo sig ptr val = do
-  case AST.typedType ptr of
-    AST.PtrTo ptrTo_ ->
-      when (AST.typedType val /= ptrTo_) $ fail $ unlines
-        [ unwords [ "Expected first value to be a pointer to some type <ty>, and"
-                  , "for the second value to be a value of type <ty>."
-                  ]
-        , "Instruction signature: " ++ sig
-        , "Pointer type:  " ++ show (AST.typedType ptr)
-        , "Value type:    " ++ show (AST.typedType val)
-        , "Pointer value: " ++ show (AST.typedValue ptr)
-        , "Value value:   " ++ show (AST.typedValue val)
-        ]
-    ty ->
-      fail $ unlines $
-        [ "Instruction expected a pointer argument."
-        , "Instruction signature: " ++ sig
-        , "Argument type: " ++ show ty
-        ]
+{-
+Note [Pointers and pointee types]
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Unlike LLVM itself, llvm-pretty and llvm-pretty-bc-parser allow mixing opaque
+and non-opaque pointers. A consequence of this is that we generally avoid
+pattern matching on PtrTo (non-opaque pointer) types and inspecting the
+underlying pointee types. This sort of code simply won't work for PtrOpaque
+types, which lack pointee types.
+
+The elimPtrTo and elimPtrTo_ functions go against this rule, as they retrieve
+the pointee type in a PtrTo. These functions are primarily used for supporting
+old versions of LLVM which do not store the necessary type information in the
+instruction itself.
+-}