[llvm] 69e8298 - [MC] eatToEndOfStatement: eat trailing EOL with Lex() (#218590)

[via llvm-commits]

Author: Fangrui Song
Date: 2026-08-25T06:55:58Z
New Revision: 69e8298ed9332fabd3cc1f2d46bfa5d215e66fe8

URL: https://github.com/llvm/llvm-project/commit/69e8298ed9332fabd3cc1f2d46bfa5d215e66fe8
DIFF: https://github.com/llvm/llvm-project/commit/69e8298ed9332fabd3cc1f2d46bfa5d215e66fe8.diff

LOG: [MC] eatToEndOfStatement: eat trailing EOL with Lex() (#218590)

eatToEndOfStatement() advances the lexer with Lexer.Lex(), which neither
filters comment tokens nor pops the include stack at Eof. It therefore
leaves the token after the newline unfiltered, and leads to a spurious
error:

```
.extern foo
/* comment */
nop

error: unexpected token at start of statement
```

Fix with Lex(). Loosely take and clean up tests from #218456

Added: 
    llvm/test/MC/AsmParser/block-comment.s

Modified: 
    llvm/lib/MC/MCParser/AsmParser.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/MC/MCParser/AsmParser.cpp b/llvm/lib/MC/MCParser/AsmParser.cpp
index c4c5b1da9aa91..320a3d896e9d0 100644
--- a/llvm/lib/MC/MCParser/AsmParser.cpp
+++ b/llvm/lib/MC/MCParser/AsmParser.cpp
@@ -1090,9 +1090,9 @@ void AsmParser::eatToEndOfStatement() {
   while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
     Lexer.Lex();
 
-  // Eat EOL.
+  // Eat EOL and skip the comments that follow.
   if (Lexer.is(AsmToken::EndOfStatement))
-    Lexer.Lex();
+    Lex();
 }
 
 StringRef AsmParser::parseStringToEndOfStatement() {

diff  --git a/llvm/test/MC/AsmParser/block-comment.s b/llvm/test/MC/AsmParser/block-comment.s
new file mode 100644
index 0000000000000..7f57936c9d6ee
--- /dev/null
+++ b/llvm/test/MC/AsmParser/block-comment.s
@@ -0,0 +1,65 @@
+## A block comment after a statement discarded by eatToEndOfStatement() is not the start of the next statement.
+# RUN: rm -rf %t && split-file %s %t && cd %t
+# RUN: llvm-mc -triple=x86_64 -preserve-comments a.s | FileCheck %s --match-full-lines --strict-whitespace
+# RUN: llvm-mc -triple=x86_64 include.s | FileCheck %s --check-prefix=INC --match-full-lines --strict-whitespace
+# RUN: not llvm-mc -triple=x86_64 err.s -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR --implicit-check-not=error:
+
+#      CHECK:	# comment{{ }}
+# CHECK-NEXT:	nop
+#CHECK-EMPTY:
+# CHECK-NEXT:	# space+comment{{ }}
+# CHECK-NEXT:	nop
+#CHECK-EMPTY:
+# CHECK-NEXT:	# spanning
+# CHECK-NEXT:	#  two lines{{ }}
+# CHECK-NEXT:	nop
+#CHECK-EMPTY:
+# CHECK-NEXT:	nop	# then an instruction{{ }}
+#CHECK-EMPTY:
+# CHECK-NEXT:	nop	# line comment
+#CHECK-EMPTY:
+# CHECK-NEXT:	# at end of file{{ }}
+
+#       INC:	retq
+# INC-EMPTY:
+#       INC:	nop
+
+#--- a.s
+.extern a
+/* comment */
+nop
+
+.extern a1
+  /* space+comment */
+nop
+
+.extern b
+/* spanning
+  two lines */
+nop
+
+.extern c
+/* then an instruction */ nop
+
+.extern d # line comment
+nop
+
+.extern e
+/* at end of file */
+
+#--- include.s
+.include "included.s"
+nop
+
+#--- included.s
+ret
+.extern b
+/* comment */
+.extern a
+#--- err.s
+.byte 1 2
+/* comment-only line */
+nop
+
+#      ERR:err.s:1:9: error: unexpected token
+# ERR-NEXT:.byte 1 2