Merge branch 'footnotes-fix-when-across-multiple-nodes' into all-footnote-fixes

Author: phillmv

src/inlines.c

@@ -1137,24 +1137,68 @@ static cmark_node *handle_close_bracket(cmark_parser *parser, subject *subj) {
   // 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] == '^') {
+
+    // 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->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;
 
-      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 *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
+      fnref->as.literal = cmark_chunk_dup(literal, 1, (fnref_end_column - fnref_start_column) - 2);
+
+      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);
+
+      // 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);
       process_emphasis(parser, subj, opener->previous_delimiter);
       pop_bracket(subj);

test/regression.txt

@@ -313,3 +313,28 @@ This is some text. It has two footnotes references, side-by-side without any spa
 </ol>
 </section>
 ````````````````````````````````
+
+Footnotes may begin with or have a 'w' or a '_' in their reference label.
+
+```````````````````````````````` example footnotes autolink
+This is some text. Sometimes the autolinker splits up text into multiple nodes, hoping it will find a hyperlink, so this text has a footnote whose reference label begins with a `w`.[^widely-cited]
+
+It has another footnote that contains many different characters (the autolinker was also breaking on `_`).[^sphinx-of-black-quartz_judge-my-vow-0123456789]
+
+[^sphinx-of-black-quartz_judge-my-vow-0123456789]: so does this.
+
+[^widely-cited]: this renders properly.
+.
+<p>This is some text. Sometimes the autolinker splits up text into multiple nodes, hoping it will find a hyperlink, so this text has a footnote whose reference label begins with a <code>w</code>.<sup class="footnote-ref"><a href="#fn1" id="fnref1">1</a></sup></p>
+<p>It has another footnote that contains many different characters (the autolinker was also breaking on <code>_</code>).<sup class="footnote-ref"><a href="#fn2" id="fnref2">2</a></sup></p>
+<section class="footnotes">
+<ol>
+<li id="fn1">
+<p>this renders properly. <a href="#fnref1" class="footnote-backref">↩</a></p>
+</li>
+<li id="fn2">
+<p>so does this. <a href="#fnref2" class="footnote-backref">↩</a></p>
+</li>
+</ol>
+</section>
+````````````````````````````````