[llvm] [MC] Fix accidentally eating the newline when handling a comment char at the end of the line. (PR #165129)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-mc

Author: Amara Emerson (aemerson)

<details>
<summary>Changes</summary>

If we have a target where both # and ## are valid comment strings,
a line ending in # would trigger the lexer to eat 2 characters
and therefore lex the *next* line as a comment. Oops. This was introduced
in 4946db15a74b761c5ac4ead18873639236b4ab5d

---
Full diff: https://github.com/llvm/llvm-project/pull/165129.diff


2 Files Affected:

- (modified) llvm/lib/MC/MCParser/AsmLexer.cpp (+8-1) 
- (added) llvm/test/MC/AsmParser/comments-x86-darwin-eol-dropped.s (+10) 


``````````diff
diff --git a/llvm/lib/MC/MCParser/AsmLexer.cpp b/llvm/lib/MC/MCParser/AsmLexer.cpp
index 968ccf776440b..8062ce88a154b 100644
--- a/llvm/lib/MC/MCParser/AsmLexer.cpp
+++ b/llvm/lib/MC/MCParser/AsmLexer.cpp
@@ -835,7 +835,14 @@ AsmToken AsmLexer::LexToken() {
   }
 
   if (isAtStartOfComment(TokStart)) {
-    CurPtr += MAI.getCommentString().size() - 1;
+    StringRef CommentString = MAI.getCommentString();
+    // For multi-char comment strings, advance CurPtr only if we matched the full
+    // string. This stops us from accidentally eating the newline if the current
+    // line ends in a single comment char.
+    if (CommentString.size() > 1 &&
+        StringRef(TokStart, CommentString.size()) == CommentString) {
+      CurPtr += CommentString.size() - 1;
+    }
     return LexLineComment();
   }
 
diff --git a/llvm/test/MC/AsmParser/comments-x86-darwin-eol-dropped.s b/llvm/test/MC/AsmParser/comments-x86-darwin-eol-dropped.s
new file mode 100644
index 0000000000000..662e5987db6e2
--- /dev/null
+++ b/llvm/test/MC/AsmParser/comments-x86-darwin-eol-dropped.s
@@ -0,0 +1,10 @@
+// RUN: llvm-mc -triple i386-apple-darwin %s 2>&1 | FileCheck %s
+.p2align 3
+// CHECK: .p2align 3
+test:
+// CHECK-LABEL: test:
+// CHECK: pushl %ebp
+// CHECK: movl %esp, %ebp
+# Check that the following line's comment # doesn't drop the movl after
+   pushl %ebp #
+   movl %esp, %ebp

``````````

</details>


https://github.com/llvm/llvm-project/pull/165129