[clang] 859d923 - Fix a failing assertion with vector type initialization

[Aaron Ballman via cfe-commits]

Author: Aaron Ballman
Date: 2022-05-04T13:25:21-04:00
New Revision: 859d92394812c0e5e650ab337717719b9687e5ce

URL: https://github.com/llvm/llvm-project/commit/859d92394812c0e5e650ab337717719b9687e5ce
DIFF: https://github.com/llvm/llvm-project/commit/859d92394812c0e5e650ab337717719b9687e5ce.diff

LOG: Fix a failing assertion with vector type initialization

When constant evaluating the initializer for an object of vector type,
we would call APInt::trunc() but truncate to the same bit-width the
object already had, which would cause an assertion. Instead, use
APInt::truncOrSelf() so that we no longer assert in this situation.

Fix #50216

Added: 
    

Modified: 
    clang/docs/ReleaseNotes.rst
    clang/lib/AST/ExprConstant.cpp
    clang/test/Sema/vector-init.c

Removed: 
    


################################################################################
diff  --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 079523a07361e..3187dad2b729e 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -132,6 +132,10 @@ Bug Fixes
   argument to a generic selection expression has side effects in the case where
   the expression is result dependent. This fixes
   `Issue 50227 <https://github.com/llvm/llvm-project/issues/50227>`_.
+- Fixed an assertion when constant evaluating an initializer for a GCC/Clang
+  floating-point vector type when the width of the initialization is exactly
+  the same as the elements of the vector being initialized.
+  Fixes `Issue 50216 <https://github.com/llvm/llvm-project/issues/50216>`_.
 
 Improvements to Clang's diagnostics
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

diff  --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp
index 70a1e0bbd91aa..650ba305c48b1 100644
--- a/clang/lib/AST/ExprConstant.cpp
+++ b/clang/lib/AST/ExprConstant.cpp
@@ -10353,9 +10353,9 @@ bool VectorExprEvaluator::VisitCastExpr(const CastExpr *E) {
       for (unsigned i = 0; i < NElts; i++) {
         llvm::APInt Elt;
         if (BigEndian)
-          Elt = SValInt.rotl(i*EltSize+FloatEltSize).trunc(FloatEltSize);
+          Elt = SValInt.rotl(i*EltSize+FloatEltSize).truncOrSelf(FloatEltSize);
         else
-          Elt = SValInt.rotr(i*EltSize).trunc(FloatEltSize);
+          Elt = SValInt.rotr(i*EltSize).truncOrSelf(FloatEltSize);
         Elts.push_back(APValue(APFloat(Sem, Elt)));
       }
     } else if (EltTy->isIntegerType()) {

diff  --git a/clang/test/Sema/vector-init.c b/clang/test/Sema/vector-init.c
index c7c278f96368f..554e976892636 100644
--- a/clang/test/Sema/vector-init.c
+++ b/clang/test/Sema/vector-init.c
@@ -42,3 +42,11 @@ void test3(void) {
   longlong2 arr1[2] = { test3_helper(), test3_helper() };
   short4 arr2[2] = { test3_helper(), test3_helper() }; // expected-error 2 {{initializing 'short4' (vector of 4 'short' values) with an expression of incompatible type 'short8' (vector of 8 'short' values)}}
 }
+
+// GH50216
+// These would previously crash when constant evaluating the initializers.
+typedef double float64x1_t __attribute__((vector_size(8)));
+float64x1_t arg1 = (float64x1_t)0x3fedf9d4343c7c80; // okay
+
+typedef float float32x1_t __attribute__((vector_size(4)));
+float32x1_t arg2 = (float32x1_t)0x3fedf9d4; // okay