markly 0.5.1 → 0.7.0

data/ext/markly/extconf.rb

@@ -1,7 +1,15 @@
 # frozen_string_literal: true
 
+# Loads mkmf which is used to make makefiles for Ruby extensions
 require 'mkmf'
 
-$CFLAGS << ' -std=c99'
+$CFLAGS << " -O3 -std=c99"
 
-create_makefile('markly/markly')
+gem_name = File.basename(__dir__)
+extension_name = 'markly'
+
+# The destination:
+dir_config(extension_name)
+
+# Generate the makefile to compile the native binary into `lib`:
+create_makefile(File.join(gem_name, extension_name))

data/ext/markly/footnotes.c

@@ -38,3 +38,26 @@ void cmark_footnote_create(cmark_map *map, cmark_node *node) {
 cmark_map *cmark_footnote_map_new(cmark_mem *mem) {
   return cmark_map_new(mem, footnote_free);
 }
+
+// Before calling `cmark_map_free` on a map with `cmark_footnotes`, first
+// unlink all of the footnote nodes before freeing their memory.
+//
+// Sometimes, two (unused) footnote nodes can end up referencing each other,
+// which as they get freed up by calling `cmark_map_free` -> `footnote_free` ->
+// etc, can lead to a use-after-free error.
+//
+// Better to `unlink` every footnote node first, setting their next, prev, and
+// parent pointers to NULL, and only then walk thru & free them up.
+void cmark_unlink_footnotes_map(cmark_map *map) {
+  cmark_map_entry *ref;
+  cmark_map_entry *next;
+
+  ref = map->refs;
+  while(ref) {
+    next = ref->next;
+    if (((cmark_footnote *)ref)->node) {
+      cmark_node_unlink(((cmark_footnote *)ref)->node);
+    }
+    ref = next;
+  }
+}

data/ext/markly/footnotes.h

@@ -18,6 +18,8 @@ typedef struct cmark_footnote cmark_footnote;
 void cmark_footnote_create(cmark_map *map, cmark_node *node);
 cmark_map *cmark_footnote_map_new(cmark_mem *mem);
 
+void cmark_unlink_footnotes_map(cmark_map *map);
+
 #ifdef __cplusplus
 }
 #endif

data/ext/markly/html.c

@@ -59,16 +59,30 @@ static void filter_html_block(cmark_html_renderer *renderer, uint8_t *data, size
     cmark_strbuf_put(html, data, (bufsize_t)len);
 }
 
-static bool S_put_footnote_backref(cmark_html_renderer *renderer, cmark_strbuf *html) {
+static bool S_put_footnote_backref(cmark_html_renderer *renderer, cmark_strbuf *html, cmark_node *node) {
   if (renderer->written_footnote_ix >= renderer->footnote_ix)
     return false;
   renderer->written_footnote_ix = renderer->footnote_ix;
 
-  cmark_strbuf_puts(html, "<a href=\"#fnref");
-  char n[32];
-  snprintf(n, sizeof(n), "%d", renderer->footnote_ix);
-  cmark_strbuf_puts(html, n);
-  cmark_strbuf_puts(html, "\" class=\"footnote-backref\">↩</a>");
+  cmark_strbuf_puts(html, "<a href=\"#fnref-");
+  houdini_escape_href(html, node->as.literal.data, node->as.literal.len);
+  cmark_strbuf_puts(html, "\" class=\"footnote-backref\" data-footnote-backref aria-label=\"Back to content\">↩</a>");
+
+  if (node->footnote.def_count > 1)
+  {
+    for(int i = 2; i <= node->footnote.def_count; i++) {
+      char n[32];
+      snprintf(n, sizeof(n), "%d", i);
+
+      cmark_strbuf_puts(html, " <a href=\"#fnref-");
+      houdini_escape_href(html, node->as.literal.data, node->as.literal.len);
+      cmark_strbuf_puts(html, "-");
+      cmark_strbuf_puts(html, n);
+      cmark_strbuf_puts(html, "\" class=\"footnote-backref\" data-footnote-backref aria-label=\"Back to content\">↩<sup class=\"footnote-ref\">");
+      cmark_strbuf_puts(html, n);
+      cmark_strbuf_puts(html, "</sup></a>");
+    }
+  }
 
   return true;
 }
@@ -273,7 +287,7 @@ static int S_render_node(cmark_html_renderer *renderer, cmark_node *node,
       } else {
         if (parent->type == CMARK_NODE_FOOTNOTE_DEFINITION && node->next == NULL) {
           cmark_strbuf_putc(html, ' ');
-          S_put_footnote_backref(renderer, html);
+          S_put_footnote_backref(renderer, html, parent);
         }
         cmark_strbuf_puts(html, "</p>\n");
       }
@@ -392,16 +406,15 @@ static int S_render_node(cmark_html_renderer *renderer, cmark_node *node,
   case CMARK_NODE_FOOTNOTE_DEFINITION:
     if (entering) {
       if (renderer->footnote_ix == 0) {
-        cmark_strbuf_puts(html, "<section class=\"footnotes\">\n<ol>\n");
+        cmark_strbuf_puts(html, "<section class=\"footnotes\" data-footnotes>\n<ol>\n");
       }
       ++renderer->footnote_ix;
-      cmark_strbuf_puts(html, "<li id=\"fn");
-      char n[32];
-      snprintf(n, sizeof(n), "%d", renderer->footnote_ix);
-      cmark_strbuf_puts(html, n);
+
+      cmark_strbuf_puts(html, "<li id=\"fn-");
+      houdini_escape_href(html, node->as.literal.data, node->as.literal.len);
       cmark_strbuf_puts(html, "\">\n");
     } else {
-      if (S_put_footnote_backref(renderer, html)) {
+      if (S_put_footnote_backref(renderer, html, node)) {
         cmark_strbuf_putc(html, '\n');
       }
       cmark_strbuf_puts(html, "</li>\n");
@@ -410,12 +423,20 @@ static int S_render_node(cmark_html_renderer *renderer, cmark_node *node,
 
   case CMARK_NODE_FOOTNOTE_REFERENCE:
     if (entering) {
-      cmark_strbuf_puts(html, "<sup class=\"footnote-ref\"><a href=\"#fn");
-      cmark_strbuf_put(html, node->as.literal.data, node->as.literal.len);
-      cmark_strbuf_puts(html, "\" id=\"fnref");
-      cmark_strbuf_put(html, node->as.literal.data, node->as.literal.len);
-      cmark_strbuf_puts(html, "\">");
-      cmark_strbuf_put(html, node->as.literal.data, node->as.literal.len);
+      cmark_strbuf_puts(html, "<sup class=\"footnote-ref\"><a href=\"#fn-");
+      houdini_escape_href(html, node->parent_footnote_def->as.literal.data, node->parent_footnote_def->as.literal.len);
+      cmark_strbuf_puts(html, "\" id=\"fnref-");
+      houdini_escape_href(html, node->parent_footnote_def->as.literal.data, node->parent_footnote_def->as.literal.len);
+
+      if (node->footnote.ref_ix > 1) {
+        char n[32];
+        snprintf(n, sizeof(n), "%d", node->footnote.ref_ix);
+        cmark_strbuf_puts(html, "-");
+        cmark_strbuf_puts(html, n);
+      }
+
+      cmark_strbuf_puts(html, "\" data-footnote-ref>");
+      houdini_escape_href(html, node->as.literal.data, node->as.literal.len);
       cmark_strbuf_puts(html, "</a></sup>");
     }
     break;

data/ext/markly/inlines.c

@@ -1137,19 +1137,77 @@ noMatch:
   // What if we're a footnote link?
   if (parser->options & CMARK_OPT_FOOTNOTES &&
       opener->inl_text->next &&
-      opener->inl_text->next->type == CMARK_NODE_TEXT &&
-      !opener->inl_text->next->next) {
+      opener->inl_text->next->type == CMARK_NODE_TEXT) {
+
     cmark_chunk *literal = &opener->inl_text->next->as.literal;
-    if (literal->len > 1 && literal->data[0] == '^') {
-      inl = make_simple(subj->mem, CMARK_NODE_FOOTNOTE_REFERENCE);
-      inl->as.literal = cmark_chunk_dup(literal, 1, literal->len - 1);
-      inl->start_line = inl->end_line = subj->line;
-      inl->start_column = opener->inl_text->start_column;
-      inl->end_column = subj->pos + subj->column_offset + subj->block_offset;
-      cmark_node_insert_before(opener->inl_text, inl);
-      cmark_node_free(opener->inl_text->next);
-      cmark_node_free(opener->inl_text);
+
+    // look back to the opening '[', and skip ahead to the next character
+    // if we're looking at a '[^' sequence, and there is other text or nodes
+    // after the ^, let's call it a footnote reference.
+    if ((literal->len > 0 && literal->data[0] == '^') && (literal->len > 1 || opener->inl_text->next->next)) {
+
+      // Before we got this far, the `handle_close_bracket` function may have
+      // advanced the current state beyond our footnote's actual closing
+      // bracket, ie if it went looking for a `link_label`.
+      // Let's just rewind the subject's position:
+      subj->pos = initial_pos;
+
+      cmark_node *fnref = make_simple(subj->mem, CMARK_NODE_FOOTNOTE_REFERENCE);
+
+      // the start and end of the footnote ref is the opening and closing brace
+      // i.e. the subject's current position, and the opener's start_column
+      int fnref_end_column = subj->pos + subj->column_offset + subj->block_offset;
+      int fnref_start_column = opener->inl_text->start_column;
+
+      // any given node delineates a substring of the line being processed,
+      // with the remainder of the line being pointed to thru its 'literal'
+      // struct member.
+      // here, we copy the literal's pointer, moving it past the '^' character
+      // for a length equal to the size of footnote reference text.
+      // i.e. end_col minus start_col, minus the [ and the ^ characters
+      //
+      // this copies the footnote reference string, even if between the
+      // `opener` and the subject's current position there are other nodes
+      //
+      // (first, check for underflows)
+      if ((fnref_start_column + 2) <= fnref_end_column) {
+        fnref->as.literal = cmark_chunk_dup(literal, 1, (fnref_end_column - fnref_start_column) - 2);
+      } else {
+        fnref->as.literal = cmark_chunk_dup(literal, 1, 0);
+      }
+
+      fnref->start_line = fnref->end_line = subj->line;
+      fnref->start_column = fnref_start_column;
+      fnref->end_column = fnref_end_column;
+
+      // we then replace the opener with this new fnref node, the net effect
+      // being replacing the opening '[' text node with a `^footnote-ref]` node.
+      cmark_node_insert_before(opener->inl_text, fnref);
+
       process_emphasis(parser, subj, opener->previous_delimiter);
+      // sometimes, the footnote reference text gets parsed into multiple nodes
+      // i.e. '[^example]' parsed into '[', '^exam', 'ple]'.
+      // this happens for ex with the autolink extension. when the autolinker
+      // finds the 'w' character, it will split the text into multiple nodes
+      // in hopes of being able to match a 'www.' substring.
+      //
+      // because this function is called one character at a time via the
+      // `parse_inlines` function, and the current subj->pos is pointing at the
+      // closing ] brace, and because we copy all the text between the [ ]
+      // braces, we should be able to safely ignore and delete any nodes after
+      // the opener->inl_text->next.
+      //
+      // therefore, here we walk thru the list and free them all up
+      cmark_node *next_node;
+      cmark_node *current_node = opener->inl_text->next;
+      while(current_node) {
+        next_node = current_node->next;
+        cmark_node_free(current_node);
+        current_node = next_node;
+      }
+
+      cmark_node_free(opener->inl_text);
+
       pop_bracket(subj);
       return NULL;
     }

data/ext/markly/markly.c

@@ -333,7 +333,7 @@ static VALUE rb_node_set_string_content(VALUE self, VALUE s) {
 static VALUE rb_node_get_type(VALUE self) {
   int node_type = 0;
   cmark_node *node = NULL;
-  VALUE symbol = NULL;
+  VALUE symbol = Qnil;
   const char *s = NULL;
 
   TypedData_Get_Struct(self, cmark_node, &rb_Markly_Node_Type, node);

data/ext/markly/node.h

@@ -76,6 +76,13 @@ struct cmark_node {
 
   cmark_syntax_extension *extension;
 
+  union {
+    int ref_ix;
+    int def_count;
+  } footnote;
+
+  cmark_node *parent_footnote_def;
+
   union {
     cmark_chunk literal;
     cmark_list list;