Parser: Introduce StrayERBClosingTagError (#1256)

This pull request introduces a new `StrayERBClosingTagError` that
detects `%>` appearing outside of ERB tags.

Inside an open tag, the `%>` is always reported as an error. The parser
consumes both tokens, adds a `LiteralNode` to preserve the source, and
continues parsing the tag. Therefore subsequent attributes and the real
`>` closing are not lost anymore:
```html+erb
<div class="foo" %> id="bar">
```

At the document top level, the `%>` is reported only in strict mode. In
non-strict mode it is silently absorbed as text, matching Erubi's
behavior for backwards compatibility, if needed.

```html+erb
<% content %> %>
```

The error message suggests using `&percnt;&gt;` HTML entities for cases
where a literal `%>` is intended.

Author: marcoroth

config.yml

@@ -307,6 +307,15 @@ errors:
         - name: opening_tag
           type: token
 
+    - name: StrayERBClosingTagError
+      message:
+        template: "Stray `%%>` found at (%u:%u). This closing delimiter is not part of an ERB tag and will be treated as plain text. If you want a literal `%%>`, use the HTML entities `%>` instead."
+        arguments:
+          - start.line
+          - start.column
+
+      fields: []
+
     - name: NestedERBTagError
       message:
         template: "ERB tag `%s` at (%u:%u) was terminated by nested `<%%` tag at (%zu:%zu). Nesting `<%%` tags is not supported."

sig/herb/errors.rbs

@@ -275,6 +275,31 @@ static AST_HTML_TEXT_NODE_T* parser_parse_text_content(parser_T* parser, hb_arra
       return NULL;
     }
 
+    if (parser->options.strict && parser->current_token->type == TOKEN_PERCENT) {
+      lexer_T lexer_copy = *parser->lexer;
+      token_T* peek_token = lexer_next_token(&lexer_copy);
+
+      if (peek_token && peek_token->type == TOKEN_HTML_TAG_END) {
+        position_T stray_start = parser->current_token->location.start;
+        position_T stray_end = peek_token->location.end;
+        token_free(peek_token);
+
+        append_strayerb_closing_tag_error(stray_start, stray_end, document_errors);
+
+        token_T* percent = parser_advance(parser);
+        hb_buffer_append(&content, percent->value);
+        token_free(percent);
+
+        token_T* gt = parser_advance(parser);
+        hb_buffer_append(&content, gt->value);
+        token_free(gt);
+
+        continue;
+      }
+
+      token_free(peek_token);
+    }
+
     token_T* token = parser_advance(parser);
     hb_buffer_append(&content, token->value);
     token_free(token);
@@ -1004,6 +1029,32 @@ static AST_HTML_OPEN_TAG_NODE_T* parser_parse_html_open_tag(parser_T* parser) {
       token_free(next_token);
     }
 
+    if (parser->current_token->type == TOKEN_PERCENT) {
+      lexer_T lexer_copy = *parser->lexer;
+      token_T* peek_token = lexer_next_token(&lexer_copy);
+
+      if (peek_token && peek_token->type == TOKEN_HTML_TAG_END) {
+        position_T stray_start = parser->current_token->location.start;
+        position_T stray_end = peek_token->location.end;
+        token_free(peek_token);
+
+        append_strayerb_closing_tag_error(stray_start, stray_end, errors);
+
+        token_T* percent = parser_advance(parser);
+        token_T* gt = parser_advance(parser);
+
+        AST_LITERAL_NODE_T* literal = ast_literal_node_init("%>", stray_start, stray_end, NULL);
+        hb_array_append(children, literal);
+
+        token_free(percent);
+        token_free(gt);
+
+        continue;
+      }
+
+      token_free(peek_token);
+    }
+
     parser_append_unexpected_error(
       parser,
       "Unexpected Token",

src/parser.c

@@ -0,0 +1,81 @@
+# frozen_string_literal: true
+
+require_relative "../test_helper"
+
+module Parser
+  class StrayERBClosingTagTest < Minitest::Spec
+    include SnapshotUtils
+
+    test "stray %> in open tag after unquoted ERB attribute value" do
+      template = <<~HTML
+        <div id=<%= dom_id(@slot) %> %>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "stray %> in open tag after quoted ERB attribute value" do
+      template = <<~HTML
+        <div id="<%= dom_id(@slot) %>" %>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "stray %> in open tag with no preceding ERB tag" do
+      template = <<~HTML
+        <div %>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "stray %> at document top level" do
+      template = <<~HTML
+        hello %> world
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "stray %> at document top level after ERB tag" do
+      template = <<~HTML
+        <% content %> %>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "%> inside quoted attribute value is not stray" do
+      template = <<~HTML
+        <div class="foo %> bar">hello</div>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "multiple stray %> in open tag" do
+      template = <<~HTML
+        <div %> %>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+
+    test "stray %> between attributes in open tag" do
+      template = <<~HTML
+        <div class="foo" %> id="bar"></div>
+      HTML
+
+      assert_parsed_snapshot(template, strict: true)
+      assert_parsed_snapshot(template, strict: false)
+    end
+  end
+end