Support parsing metadata fields added in LLVM 22

.github/workflows/nix-ci.yml

@@ -7,6 +7,10 @@ on:
   pull_request:
   workflow_dispatch:
 
+defaults:
+  run:
+    shell: bash
+
 jobs:
   build:
     # This job builds the main package, via a number of GHC versions.  This is
@@ -32,7 +36,6 @@ jobs:
           name: galois
           authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
       - name: build ${{ matrix.ghc-version }}
-        shell: bash
         run: nix build -L github:${{ github.repository }}/${{ github.sha }}#llvm-pretty-bc-parser.${{ matrix.ghc-version }}
 
   build_old_GHC:
@@ -63,13 +66,12 @@ jobs:
           name: galois
           authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
       - name: build ${{ matrix.ghc-version }}
-        shell: bash
         run: |
           nix build -L \
           github:${{ github.repository }}/${{ github.sha }}#llvm-pretty-bc-parser.${{ matrix.ghc-version }} \
           --override-input nixpkgs github:nixos/nixpkgs/25.05
 
-  tests_llvm16-21:
+  tests_llvm22:
     # This job runs the LLVM tests.  The builds above verify that the code can be
     # built with different versions of GHC.  However, the resulting code should
     # be able to parse and represent the bitcode from a wide range of LLVM
@@ -94,11 +96,26 @@ jobs:
           name: galois
           authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
       - name: test
-        shell: bash
+        run: nix build -L github:${{ github.repository }}/${{ github.sha }}#TESTS_22
+
+  tests_llvm16-21:
+    # See note for tests_llvm22 above.  This job is identical except for the
+    # nix build target.
+    runs-on: ubuntu-latest
+    steps:
+      - uses: cachix/install-nix-action@v20
+        with:
+          nix_path: nixpkgs=channel:nixos-23.05
+          github_access_token: ${{ secrets.GITHUB_TOKEN }}
+      - uses: cachix/cachix-action@v15
+        with:
+          name: galois
+          authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
+      - name: test
         run: nix build -L github:${{ github.repository }}/${{ github.sha }}#TESTS_16-21
 
   tests_llvm10-15:
-    # See note for tests_llvm16-21 above.  This job is identical except for the
+    # See note for tests_llvm22 above.  This job is identical except for the
     # nix build target.
     runs-on: ubuntu-latest
     steps:
@@ -111,5 +128,4 @@ jobs:
           name: galois
           authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
       - name: test
-        shell: bash
         run: nix build -L github:${{ github.repository }}/${{ github.sha }}#TESTS_10-15

CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## next
 
+* Support LLVM 22:
+  * Support parsing `sourceLanguageVersion` fields in `DICompileUnit` debug
+    metadata.
+  * Support parsing `dataSize` fields in `DIBasicType` debug metadata.
 * Add support for reading 16-bit "Brain" floats (`bfloat`)
 * Add missing handling for 16-bit IEEE half-precision floats,
   64-bit IEEE quad-precision floats, and PowerPC pair-of-doubles

README.md

@@ -39,6 +39,7 @@ compilers.
 |           | v19     | ✓                             | ✓            | See [issues][llvm19]  |
 |           | v20     | ✓                             | ✓            | See [issues][llvm20]  |
 |           | v21     | ✓                             | ✓            | See [issues][llvm21]  |
+|           | v22     | ✓                             | ✓            | See [issues][llvm22]  |
 
 If you encounter problems with the output of *any* compiler, please file [an
 issue](https://github.com/GaloisInc/llvm-pretty-bc-parser/issues).
@@ -82,3 +83,4 @@ the developers' documentation for more details and a rationale:
 [llvm19]: https://github.com/GaloisInc/llvm-pretty-bc-parser/issues?q=is%3Aopen+is%3Aissue+label%3Allvm%2F19.0
 [llvm20]: https://github.com/GaloisInc/llvm-pretty-bc-parser/issues?q=is%3Aopen+is%3Aissue+label%3Allvm%2F20.0
 [llvm21]: https://github.com/GaloisInc/llvm-pretty-bc-parser/issues?q=is%3Aopen+is%3Aissue+label%3Allvm%2F21.0
+[llvm22]: https://github.com/GaloisInc/llvm-pretty-bc-parser/issues?q=is%3Aopen+is%3Aissue+label%3Allvm%2F22.0

flake.nix

@@ -41,8 +41,11 @@
           # (https://gitlab.haskell.org/ghc/ghc/-/issues/25771).  This bug is
           # fixed in GHC 9.12.3; remove the following ghcver setting when GHC
           # 9.12.3 is available in nixpkgs.
+          #
+          # GHC 9.14 dependencies are not ready yet.
           builtins.filter
-            (n: builtins.substring 0 6 n != "ghc912")
+            (n: builtins.substring 0 6 n != "ghc912" &&
+                builtins.substring 0 6 n != "ghc914")
             (levers.validGHCVersions pkgs.haskell.compiler);
     in
     rec {
@@ -59,8 +62,8 @@
             # nixpkgs_old_llvm.legacyPackages.x86_64-linux.clang_16
             # nixpkgs_old_llvm.legacyPackages.x86_64-linux.llvm_16
 
-            nixpkgs.legacyPackages.x86_64-linux.clang_21
-            nixpkgs.legacyPackages.x86_64-linux.llvm_21
+            nixpkgs.legacyPackages.x86_64-linux.clang_22
+            nixpkgs.legacyPackages.x86_64-linux.llvm_22
 
             # Other packages to add to the development shell:
             pkgs.cabal-install
@@ -136,6 +139,7 @@
               # in separate jobs to avoid this.
               TESTS_10-15
               TESTS_16-21
+              TESTS_22
             ];
           TESTS_10-15 = wrap "llvm-pretty-bc-parser-TESTS_10-15"
             (builtins.map
@@ -161,6 +165,13 @@
                 "21"
               ]
             );
+          TESTS_22 = wrap "llvm-pretty-bc-parser-TESTS_22"
+            (builtins.map
+              (llvm-pretty-bc-parser-test llvm-pretty-bc-parser)
+              [
+                "22"
+              ]
+            );
           llvm-pretty = mkHaskell "llvm-pretty" llvm-pretty-src {};
           llvm-pretty-bc-parser = mkHaskell "llvm-pretty-bc-parser" self {
             inherit llvm-pretty tasty-sugar;

src/Data/LLVM/BitCode/IR/Metadata.hs

@@ -46,7 +46,7 @@
 import           Data.Either (partitionEithers)
 import           Data.Generics.Uniplate.Data
 import qualified Data.IntMap as IntMap
 import           Data.List (mapAccumL, foldl')
 import           Data.Map (Map)
 import qualified Data.Map as Map
 import           Data.Maybe (fromMaybe, mapMaybe)
@@ -572,7 +572,7 @@
       return $! updateMetadataTable (addDebugInfo isDistinct diEnum) pm
 
     15 -> label "METADATA_BASIC_TYPE" $ do
-      assertRecordSizeBetween r 6 8
+      assertRecordSizeBetween r 6 9
       ctx        <- getContext
       flags      <- parseField r 0 numeric
       let isDistinct     = testBit (flags :: Int) 0
@@ -589,6 +589,10 @@
         if length (recordFields r) <= 7
         then pure 0
         else parseField r 7 numeric
+      dibtDataSize <-
+        if length (recordFields r) <= 8
+        then pure 0
+        else parseField r 8 numeric
       let dibt = DIBasicType {..}
       return $! updateMetadataTable
         (addDebugInfo isDistinct (DebugInfoBasicType dibt)) pm
@@ -697,7 +701,7 @@
         (addDebugInfo isDistinct (DebugInfoSubroutineType dist)) pm
 
     20 -> label "METADATA_COMPILE_UNIT" $ do
-      assertRecordSizeBetween r 14 22
+      assertRecordSizeBetween r 14 23
       let recordSize = length (recordFields r)
       ctx        <- getContext
       isDistinct <- parseField r 0 nonzero
@@ -738,6 +742,10 @@
       dicuSDK <- if recordSize <= 21
                  then pure Nothing
                  else mdStringOrNull ctx pm <$> parseField r 21 numeric
+      dicuSourceLanguageVersion <-
+        if recordSize <= 22
+        then pure 0
+        else parseField r 22 numeric
       let dicu = DICompileUnit {..}
       return $! updateMetadataTable
         (addDebugInfo isDistinct (DebugInfoCompileUnit dicu)) pm