From e78398554d27be2dc1a2c366571ee011989b5ce0 Mon Sep 17 00:00:00 2001
From: Kevin Newton <kddnewton@gmail.com>
Date: Wed, 1 Nov 2023 14:28:01 -0400
Subject: [PATCH] Handle invalid constants in the constant pool

It's possible for us to parse a constant that is invalid in the
current encoding. To fix this, we wrap the intern function in an
rb_protect to ensure we always put a valid symbol into the AST.
---
 Makefile                           |  6 ++++++
 src/prism.c                        |  7 +++++--
 templates/ext/prism/api_node.c.erb | 12 +++++++++++-
 3 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/Makefile b/Makefile
index dd08ce3b616..4a0fbc659a2 100644
--- a/Makefile
+++ b/Makefile
@@ -88,3 +88,9 @@ clean:
 all-no-debug: DEBUG_FLAGS := -DNDEBUG=1
 all-no-debug: OPTFLAGS := -O3
 all-no-debug: all
+
+run: Makefile $(STATIC_OBJECTS) $(HEADERS) test.c
+	$(ECHO) "compiling test.c"
+	$(Q) $(CC) $(CPPFLAGS) $(CFLAGS) $(STATIC_OBJECTS) test.c
+	$(ECHO) "running test.c"
+	$(Q) ./a.out
diff --git a/src/prism.c b/src/prism.c
index 42d97120634..de32ccae63e 100644
--- a/src/prism.c
+++ b/src/prism.c
@@ -8633,13 +8633,16 @@ parser_lex(pm_parser_t *parser) {
                     if (!lex_state_beg_p(parser) && match(parser, '=')) {
                         lex_state_set(parser, PM_LEX_STATE_BEG);
                         LEX(PM_TOKEN_PERCENT_EQUAL);
-                    }
-                    else if(
+                    } else if (
                         lex_state_beg_p(parser) ||
                         (lex_state_p(parser, PM_LEX_STATE_FITEM) && (peek(parser) == 's')) ||
                         lex_state_spcarg_p(parser, space_seen)
                     ) {
                         if (!parser->encoding.alnum_char(parser->current.end, parser->end - parser->current.end)) {
+                            if (*parser->current.end >= 0x80) {
+                                pm_parser_err_current(parser, PM_ERR_INVALID_PERCENT);
+                            }
+
                             lex_mode_push_string(parser, true, false, lex_mode_incrementor(*parser->current.end), lex_mode_terminator(*parser->current.end));
 
                             size_t eol_length = match_eol(parser);
diff --git a/templates/ext/prism/api_node.c.erb b/templates/ext/prism/api_node.c.erb
index 6da58321855..1ffbf6c9f84 100644
--- a/templates/ext/prism/api_node.c.erb
+++ b/templates/ext/prism/api_node.c.erb
@@ -83,7 +83,17 @@ pm_ast_new(pm_parser_t *parser, pm_node_t *node, rb_encoding *encoding) {
 
     for (uint32_t index = 0; index < parser->constant_pool.size; index++) {
         pm_constant_t *constant = &parser->constant_pool.constants[index];
-        constants[index] = rb_intern3((const char *) constant->start, constant->length, encoding);
+        int state = 0;
+
+        VALUE string = rb_enc_str_new((const char *) constant->start, constant->length, encoding);
+        ID value = rb_protect(rb_intern_str, string, &state);
+
+        if (state != 0) {
+            value = rb_intern_const("?");
+            rb_set_errinfo(Qnil);
+        }
+
+        constants[index] = value;
     }
 
     pm_node_stack_node_t *node_stack = NULL;