[cpp] fix Semgrep pattern parsing for class/struct/enum bodies and operator overloads

fyi/semgrep-grammars/src/semgrep-cpp/grammar.js

@@ -33,7 +33,10 @@ module.exports = grammar(base_grammar, {
 
     // Typed metavariables
 
-    semgrep_metavar: $ => /\$[A-Z_][A-Z_0-9]*/,
+    // Token precedence so the lexer prefers `semgrep_metavar` over the
+    // `identifier` rule (whose regex also accepts a leading `$`) when
+    // the surrounding rule allows a metavar (e.g. inside an enum body).
+    semgrep_metavar: $ => token(prec(1, /\$[A-Z_][A-Z_0-9]*/)),
 
     semgrep_typed_metavar: $ =>
       seq(
@@ -99,6 +102,50 @@ module.exports = grammar(base_grammar, {
       $.semgrep_ellipsis
     ),
 
+    // Allow `...` and `$X` inside class/struct bodies, e.g.
+    //   class C { ... }
+    //   struct S { $X }
+    // The lexer prefers `semgrep_metavar` over `identifier` thanks to
+    // the token precedence on `semgrep_metavar`, so a bare `$X` ends
+    // up here rather than as a malformed field_declaration.
+    _field_declaration_list_item: ($, previous) => choice(
+      previous,
+      $.semgrep_ellipsis,
+      $.semgrep_metavar
+    ),
+
+    // Allow `...` and `$X` inside enum bodies, e.g.
+    //   enum E { ... }
+    // The upstream rule is a `seq` so we must override it wholesale
+    // (copied from tree-sitter-c with semgrep alternatives added).
+    enumerator_list: $ => seq(
+      '{',
+      repeat(choice(
+        seq($.enumerator, ','),
+        alias($.preproc_if_in_enumerator_list, $.preproc_if),
+        alias($.preproc_ifdef_in_enumerator_list, $.preproc_ifdef),
+        seq($.preproc_call, ','),
+        seq($.semgrep_ellipsis, ','),
+        seq($.semgrep_metavar, ','),
+      )),
+      optional(choice(
+        $.enumerator,
+        alias($.preproc_if_in_enumerator_list_no_comma, $.preproc_if),
+        alias($.preproc_ifdef_in_enumerator_list_no_comma, $.preproc_ifdef),
+        $.preproc_call,
+        $.semgrep_ellipsis,
+        $.semgrep_metavar,
+      )),
+      '}',
+    ),
+
+    // Allow `operator $OP` as an operator name, e.g.
+    //   T operator $OP(T x);
+    operator_name: ($, previous) => choice(
+      previous,
+      prec(1, seq('operator', $.semgrep_metavar)),
+    ),
+
     // So we prefer to parse a unary left fold for
     // 1 + ...
     // rather than the addition of an ellipsis

fyi/versions

@@ -2,11 +2,11 @@ File: semgrep-grammars/src/tree-sitter-c/LICENSE
 Git repo name: tree-sitter-c
 Latest commit in repo: ecdd500806cf8154d944344f1df6418b32e0e9a7
 Last change in file:
-  commit ecdd500806cf8154d944344f1df6418b32e0e9a7
-  Author: Brandon Wu <49291449+brandonspark@users.noreply.github.com>
-  Date:   Fri Jan 26 12:20:17 2024 -0800
+  commit af2904ca831cbb33cdc3bb7bacc2ed9840bd0a3c
+  Author: Max Brunsfeld <maxbrunsfeld@gmail.com>
+  Date:   Fri Sep 5 23:53:51 2014 -0700
 
-      fix: properly suffix elifdef
+      Initial commit
 ---
 File: semgrep-grammars/src/tree-sitter-c/grammar.js
 Git repo name: tree-sitter-c
@@ -22,11 +22,11 @@ File: semgrep-grammars/src/tree-sitter-cpp/LICENSE
 Git repo name: tree-sitter-cpp
 Latest commit in repo: 4ca37be8e70e5a40ae95688bec56b886ba945888
 Last change in file:
-  commit 4ca37be8e70e5a40ae95688bec56b886ba945888
-  Author: Brandon Wu <49291449+brandonspark@users.noreply.github.com>
-  Date:   Fri Jan 26 08:33:26 2024 -0800
+  commit d8966822015417d060ffaafd95a71a49dd28ec99
+  Author: Max Brunsfeld <maxbrunsfeld@gmail.com>
+  Date:   Fri Jan 15 11:04:19 2016 -0800
 
-      disambiguate fold vs parenthesized assignment (#239)
+      Initial commit
 ---
 File: semgrep-grammars/src/tree-sitter-cpp/grammar.js
 Git repo name: tree-sitter-cpp
@@ -39,12 +39,30 @@ Last change in file:
       disambiguate fold vs parenthesized assignment (#239)
 ---
 File: semgrep-grammars/src/semgrep-cpp/grammar.js
-Git repo name: ocaml-tree-sitter-semgrep
-Latest commit in repo: 091f5438fc0c15b80217f00e5b94ec0e55517383
+Git repo name: agent-aab45847a71009554
+Latest commit in repo: 9244700fc8ed0c101f640a33c6f4c115ad73d3b5
 Last change in file:
-  commit 5f3836d894376e97f7582fd32c2bd01b98697886
-  Author: brandonspark <wu.brandonj@gmail.com>
-  Date:   Mon Jan 29 15:03:38 2024 -0800
+  commit 9244700fc8ed0c101f640a33c6f4c115ad73d3b5
+  Author: brandonspark <brandon@semgrep.com>
+  Date:   Wed Apr 29 17:34:00 2026 -0700
 
-      remove redundant choice
+      fix(cpp): allow Semgrep patterns in class/struct/enum bodies and operator overloads
+
+      Augments the C++ grammar so canonical Semgrep patterns parse cleanly:
+
+      - LANG-486: `...` and `$X` inside class/struct bodies (via
+        `_field_declaration_list_item`) and enum bodies (via a wholesale
+        override of `enumerator_list`, since the upstream rule is a `seq`).
+      - LANG-497: `operator $OP` as an operator name (via a `previous`-style
+        augmentation of `operator_name`).
+
+      The `semgrep_metavar` token gets a precedence bump so the lexer prefers
+      it over `identifier` (whose regex also accepts a leading `$`) when both
+      are valid at a position; this is what lets a bare `$X` parse as a
+      metavariable inside enum/struct bodies instead of as a malformed
+      declaration or enumerator.
+
+      Adds 6 corpus tests covering the minimal ticket repros.
+
+      Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 ---