[llvm] 25752a4 - [X86AsmParser] IntelExpression: End of Statement should check for valid end state (#95677)

[via llvm-commits]

Author: v01dXYZ
Date: 2024-07-16T09:30:19+08:00
New Revision: 25752a40b8b224e3695edd506f304866f065ca9a

URL: https://github.com/llvm/llvm-project/commit/25752a40b8b224e3695edd506f304866f065ca9a
DIFF: https://github.com/llvm/llvm-project/commit/25752a40b8b224e3695edd506f304866f065ca9a.diff

LOG: [X86AsmParser] IntelExpression: End of Statement should check for valid end state (#95677)

The following commit bfb7099eeb9b6f62510b1db0cb93a8c3cfa68236 added a
special case for End of Statement that doesn't check if the state
machine is rightfully in a state where ending is valid.

This PR suggest to revert this change to make `EndOfStatement` processed
as any other tokens that are not consumable by the state machine.

Fixes https://github.com/llvm/llvm-project/issues/94446

---------

Co-authored-by: v01dxyz <v01dxyz at v01d.xyz>

Added: 
    llvm/test/MC/X86/intel-syntax-expr.s

Modified: 
    llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
index e49e96ceef6a4..daf982322bce9 100644
--- a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -475,7 +475,9 @@ class X86AsmParser : public MCTargetAsmParser {
     unsigned getLength() const { return CurType.Length; }
     int64_t getImm() { return Imm + IC.execute(); }
     bool isValidEndState() const {
-      return State == IES_RBRAC || State == IES_INTEGER;
+      return State == IES_RBRAC || State == IES_RPAREN ||
+             State == IES_INTEGER || State == IES_REGISTER ||
+             State == IES_OFFSET;
     }
 
     // Is the intel expression appended after an operand index.
@@ -1898,9 +1900,6 @@ bool X86AsmParser::ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End) {
     case AsmToken::Error:
       return Error(getLexer().getErrLoc(), getLexer().getErr());
       break;
-    case AsmToken::EndOfStatement:
-      Done = true;
-      break;
     case AsmToken::Real:
       // DotOperator: [ebx].0
       UpdateLocLex = false;

diff  --git a/llvm/test/MC/X86/intel-syntax-expr.s b/llvm/test/MC/X86/intel-syntax-expr.s
new file mode 100644
index 0000000000000..8aa083107dd85
--- /dev/null
+++ b/llvm/test/MC/X86/intel-syntax-expr.s
@@ -0,0 +1,8 @@
+// RUN: not llvm-mc -triple x86_64-unknown-unknown -x86-asm-syntax=intel %s 2>&1 | FileCheck %s
+
+// When the intel syntax is enabled, to parse an operand, X86AsmParser doesn't use the method parseExpression from AsmParser
+// but ParseIntelExpr which was not processing well an end of statement.
+
+// CHECK: error: unknown token in expression
+test:
+i-