[ruby/prism] Convert start line to signed integers

Ruby allows for 0 or negative line start, this is often used
with `eval` calls to get a correct offset when prefixing a snippet.

e.g.

```ruby
caller = caller_locations(1, 1).first
class_eval <<~RUBY, caller.path, caller.line - 2
  # frozen_string_literal: true
  def some_method
    #{caller_provided_code_snippet}
  end
RUBY
```

https://github.com/ruby/prism/commit/0d14ed1452

11 files changed, 74 insertions, 18 deletions

diff --git a/prism/extension.c b/prism/extension.c
index 3637cc1617..c3ee58d15e 100644
--- a/prism/extension.c
+++ b/prism/extension.c
@@ -126,7 +126,7 @@ build_options_i(VALUE key, VALUE value, VALUE argument) {
     } else if (key_id == rb_option_id_encoding) {
         if (!NIL_P(value)) pm_options_encoding_set(options, rb_enc_name(rb_to_encoding(value)));
     } else if (key_id == rb_option_id_line) {
-        if (!NIL_P(value)) pm_options_line_set(options, NUM2UINT(value));
+        if (!NIL_P(value)) pm_options_line_set(options, NUM2INT(value));
     } else if (key_id == rb_option_id_frozen_string_literal) {
         if (!NIL_P(value)) pm_options_frozen_string_literal_set(options, value == Qtrue);
     } else if (key_id == rb_option_id_verbose) {
@@ -166,6 +166,7 @@ build_options(VALUE argument) {
  */
 static void
 extract_options(pm_options_t *options, VALUE filepath, VALUE keywords) {
+    options->line = 1; // default
     if (!NIL_P(keywords)) {
         struct build_options_data data = { .options = options, .keywords = keywords };
         struct build_options_data *argument = &data;
diff --git a/prism/options.c b/prism/options.c
index 84c1fcbb39..85d04d6272 100644
--- a/prism/options.c
+++ b/prism/options.c
@@ -20,7 +20,7 @@ pm_options_encoding_set(pm_options_t *options, const char *encoding) {
  * Set the line option on the given options struct.
  */
 PRISM_EXPORTED_FUNCTION void
-pm_options_line_set(pm_options_t *options, uint32_t line) {
+pm_options_line_set(pm_options_t *options, int32_t line) {
     options->line = line;
 }
 
@@ -115,6 +115,22 @@ pm_options_read_u32(const char *data) {
 }
 
 /**
+ * Read a 32-bit signed integer from a pointer. This function is used to read
+ * the options that are passed into the parser from the Ruby implementation. It
+ * handles aligned and unaligned reads.
+ */
+static int32_t
+pm_options_read_s32(const char *data) {
+    if (((uintptr_t) data) % sizeof(int32_t) == 0) {
+        return *((int32_t *) data);
+    } else {
+        int32_t value;
+        memcpy(&value, data, sizeof(int32_t));
+        return value;
+    }
+}
+
+/**
  * Deserialize an options struct from the given binary string. This is used to
  * pass options to the parser from an FFI call so that consumers of the library
  * from an FFI perspective don't have to worry about the structure of our
@@ -123,6 +139,9 @@ pm_options_read_u32(const char *data) {
  */
 void
 pm_options_read(pm_options_t *options, const char *data) {
+    options->line = 1; // default
+    if (data == NULL) return;
+
     uint32_t filepath_length = pm_options_read_u32(data);
     data += 4;
 
@@ -131,7 +150,7 @@ pm_options_read(pm_options_t *options, const char *data) {
         data += filepath_length;
     }
 
-    options->line = pm_options_read_u32(data);
+    options->line = pm_options_read_s32(data);
     data += 4;
 
     uint32_t encoding_length = pm_options_read_u32(data);
diff --git a/prism/options.h b/prism/options.h
index 2ea85c838c..8608838da8 100644
--- a/prism/options.h
+++ b/prism/options.h
@@ -35,7 +35,7 @@ typedef struct {
      * The line within the file that the parse starts on. This value is
      * 0-indexed.
      */
-    uint32_t line;
+    int32_t line;
 
     /**
      * The name of the encoding that the source file is in. Note that this must
@@ -80,7 +80,7 @@ PRISM_EXPORTED_FUNCTION void pm_options_filepath_set(pm_options_t *options, cons
  * @param options The options struct to set the line on.
  * @param line The line to set.
  */
-PRISM_EXPORTED_FUNCTION void pm_options_line_set(pm_options_t *options, uint32_t line);
+PRISM_EXPORTED_FUNCTION void pm_options_line_set(pm_options_t *options, int32_t line);
 
 /**
  * Set the encoding option on the given options struct.
diff --git a/prism/parser.h b/prism/parser.h
index c1f9e0f663..dfc15e19b7 100644
--- a/prism/parser.h
+++ b/prism/parser.h
@@ -661,7 +661,7 @@ struct pm_parser {
      * The line number at the start of the parse. This will be used to offset
      * the line numbers of all of the locations.
      */
-    uint32_t start_line;
+    int32_t start_line;
 
     /** Whether or not we're at the beginning of a command. */
     bool command_start;
diff --git a/prism/prism.c b/prism/prism.c
index aee9fc7b88..1a4d31c268 100644
--- a/prism/prism.c
+++ b/prism/prism.c
@@ -17067,9 +17067,7 @@ pm_parser_init(pm_parser_t *parser, const uint8_t *source, size_t size, const pm
         parser->filepath_string = options->filepath;
 
         // line option
-        if (options->line > 0) {
-            parser->start_line = options->line;
-        }
+        parser->start_line = options->line;
 
         // encoding option
         size_t encoding_length = pm_string_length(&options->encoding);
@@ -17238,7 +17236,7 @@ pm_serialize(pm_parser_t *parser, pm_node_t *node, pm_buffer_t *buffer) {
 PRISM_EXPORTED_FUNCTION void
 pm_serialize_parse(pm_buffer_t *buffer, const uint8_t *source, size_t size, const char *data) {
     pm_options_t options = { 0 };
-    if (data != NULL) pm_options_read(&options, data);
+    pm_options_read(&options, data);
 
     pm_parser_t parser;
     pm_parser_init(&parser, source, size, &options);
@@ -17260,7 +17258,7 @@ pm_serialize_parse(pm_buffer_t *buffer, const uint8_t *source, size_t size, cons
 PRISM_EXPORTED_FUNCTION void
 pm_serialize_parse_comments(pm_buffer_t *buffer, const uint8_t *source, size_t size, const char *data) {
     pm_options_t options = { 0 };
-    if (data != NULL) pm_options_read(&options, data);
+    pm_options_read(&options, data);
 
     pm_parser_t parser;
     pm_parser_init(&parser, source, size, &options);
@@ -17268,7 +17266,7 @@ pm_serialize_parse_comments(pm_buffer_t *buffer, const uint8_t *source, size_t s
     pm_node_t *node = pm_parse(&parser);
     pm_serialize_header(buffer);
     pm_serialize_encoding(&parser.encoding, buffer);
-    pm_buffer_append_varuint(buffer, parser.start_line);
+    pm_buffer_append_varsint(buffer, parser.start_line);
     pm_serialize_comment_list(&parser, &parser.comment_list, buffer);
 
     pm_node_destroy(&parser, node);
diff --git a/prism/templates/ext/prism/api_node.c.erb b/prism/templates/ext/prism/api_node.c.erb
index 7bc52c1120..5811cf2027 100644
--- a/prism/templates/ext/prism/api_node.c.erb
+++ b/prism/templates/ext/prism/api_node.c.erb
@@ -46,7 +46,7 @@ pm_source_new(pm_parser_t *parser, rb_encoding *encoding) {
         rb_ary_push(offsets, INT2FIX(parser->newline_list.offsets[index]));
     }
 
-    VALUE source_argv[] = { source, ULONG2NUM(parser->start_line), offsets };
+    VALUE source_argv[] = { source, LONG2NUM(parser->start_line), offsets };
     return rb_class_new_instance(3, source_argv, rb_cPrismSource);
 }
 
diff --git a/prism/templates/lib/prism/serialize.rb.erb b/prism/templates/lib/prism/serialize.rb.erb
index a38c796971..350a502d6a 100644
--- a/prism/templates/lib/prism/serialize.rb.erb
+++ b/prism/templates/lib/prism/serialize.rb.erb
@@ -79,7 +79,7 @@ module Prism
       end
 
       def load_start_line
-        source.start_line = load_varuint
+        source.start_line = load_varsint
       end
 
       def load_comments
@@ -161,6 +161,11 @@ module Prism
         end
       end
 
+      def load_varsint
+        n = load_varuint
+        (n >> 1) ^ (-(n & 1))
+      end
+
       def load_serialized_length
         io.read(4).unpack1("L")
       end
diff --git a/prism/templates/src/serialize.c.erb b/prism/templates/src/serialize.c.erb
index 0f6b119d5b..e82a8703b2 100644
--- a/prism/templates/src/serialize.c.erb
+++ b/prism/templates/src/serialize.c.erb
@@ -219,7 +219,7 @@ pm_serialize_encoding(pm_encoding_t *encoding, pm_buffer_t *buffer) {
 void
 pm_serialize_content(pm_parser_t *parser, pm_node_t *node, pm_buffer_t *buffer) {
     pm_serialize_encoding(&parser->encoding, buffer);
-    pm_buffer_append_varuint(buffer, parser->start_line);
+    pm_buffer_append_varsint(buffer, parser->start_line);
 <%- unless Prism::SERIALIZE_ONLY_SEMANTICS_FIELDS -%>
     pm_serialize_comment_list(parser, &parser->comment_list, buffer);
 <%- end -%>
@@ -301,7 +301,7 @@ serialize_token(void *data, pm_parser_t *parser, pm_token_t *token) {
 PRISM_EXPORTED_FUNCTION void
 pm_serialize_lex(pm_buffer_t *buffer, const uint8_t *source, size_t size, const char *data) {
     pm_options_t options = { 0 };
-    if (data != NULL) pm_options_read(&options, data);
+    pm_options_read(&options, data);
 
     pm_parser_t parser;
     pm_parser_init(&parser, source, size, &options);
@@ -318,7 +318,7 @@ pm_serialize_lex(pm_buffer_t *buffer, const uint8_t *source, size_t size, const
     pm_buffer_append_byte(buffer, 0);
 
     pm_serialize_encoding(&parser.encoding, buffer);
-    pm_buffer_append_varuint(buffer, parser.start_line);
+    pm_buffer_append_varsint(buffer, parser.start_line);
     pm_serialize_comment_list(&parser, &parser.comment_list, buffer);
     pm_serialize_magic_comment_list(&parser, &parser.magic_comment_list, buffer);
     pm_serialize_data_loc(&parser, buffer);
@@ -337,7 +337,7 @@ pm_serialize_lex(pm_buffer_t *buffer, const uint8_t *source, size_t size, const
 PRISM_EXPORTED_FUNCTION void
 pm_serialize_parse_lex(pm_buffer_t *buffer, const uint8_t *source, size_t size, const char *data) {
     pm_options_t options = { 0 };
-    if (data != NULL) pm_options_read(&options, data);
+    pm_options_read(&options, data);
 
     pm_parser_t parser;
     pm_parser_init(&parser, source, size, &options);
diff --git a/prism/util/pm_buffer.c b/prism/util/pm_buffer.c
index dcdf1770bb..307b55d030 100644
--- a/prism/util/pm_buffer.c
+++ b/prism/util/pm_buffer.c
@@ -152,6 +152,15 @@ pm_buffer_append_varuint(pm_buffer_t *buffer, uint32_t value) {
 }
 
 /**
+ * Append a 32-bit signed integer to the buffer as a variable-length integer.
+ */
+void
+pm_buffer_append_varsint(pm_buffer_t *buffer, int32_t value) {
+    uint32_t unsigned_int = ((uint32_t)(value) << 1) ^ ((uint32_t)(value >> 31));
+    pm_buffer_append_varuint(buffer, unsigned_int);
+}
+
+/**
  * Concatenate one buffer onto another.
  */
 void
diff --git a/prism/util/pm_buffer.h b/prism/util/pm_buffer.h
index a8596be476..ec11d05e9b 100644
--- a/prism/util/pm_buffer.h
+++ b/prism/util/pm_buffer.h
@@ -121,6 +121,14 @@ void pm_buffer_append_byte(pm_buffer_t *buffer, uint8_t value);
 void pm_buffer_append_varuint(pm_buffer_t *buffer, uint32_t value);
 
 /**
+ * Append a 32-bit signed integer to the buffer as a variable-length integer.
+ *
+ * @param buffer The buffer to append to.
+ * @param value The integer to append.
+ */
+void pm_buffer_append_varsint(pm_buffer_t *buffer, int32_t value);
+
+/**
  * Concatenate one buffer onto another.
  *
  * @param destination The buffer to concatenate onto.
diff --git a/test/prism/parse_test.rb b/test/prism/parse_test.rb
index 6bd7a5d2a1..2feb15b48b 100644
--- a/test/prism/parse_test.rb
+++ b/test/prism/parse_test.rb
@@ -46,6 +46,22 @@ module Prism
       assert_equal filepath, find_source_file_node(result.value).filepath
     end
 
+    def test_parse_takes_line
+      line = 4
+      result = Prism.parse("def foo\n __FILE__\nend", line: line)
+
+      assert_equal line, result.value.location.start_line
+      assert_equal line + 1, find_source_file_node(result.value).location.start_line
+    end
+
+    def test_parse_takes_negative_lines
+      line = -2
+      result = Prism.parse("def foo\n __FILE__\nend", line: line)
+
+      assert_equal line, result.value.location.start_line
+      assert_equal line + 1, find_source_file_node(result.value).location.start_line
+    end
+
     def test_parse_lex
       node, tokens = Prism.parse_lex("def foo; end").value