Merge pull request #2650 from Elberet/fix-parser

Fix: block parser uses regexes for stop tags, allows stricter matching

app/assets/javascripts/discourse/dialects/bbcode_dialect.js

@@ -120,7 +120,7 @@ Discourse.Markdown.whiteListTag('span', 'class', /^bbcode-size-\d+$/);
 // Handles `[code] ... [/code]` blocks
 Discourse.Dialect.replaceBlock({
   start: /(\[code\])([\s\S]*)/igm,
-  stop: '[/code]',
+  stop: /\[\/code\]/igm,
   rawContents: true,
 
   emitter: function(blockContents) {

app/assets/javascripts/discourse/dialects/code_dialect.js

@@ -21,7 +21,7 @@ function flattenBlocks(blocks) {
 
 Discourse.Dialect.replaceBlock({
   start: /^`{3}([^\n\[\]]+)?\n?([\s\S]*)?/gm,
-  stop: '```',
+  stop: /^```$/gm,
   emitter: function(blockContents, matches) {
 
     var klass = Discourse.SiteSettings.default_code_lang;
@@ -54,7 +54,7 @@ Discourse.Dialect.on('parseNode', function (event) {
 
 Discourse.Dialect.replaceBlock({
   start: /(<pre[^\>]*\>)([\s\S]*)/igm,
-  stop: '</pre>',
+  stop: /<\/pre>/igm,
   rawContents: true,
   skipIfTradtionalLinebreaks: true,
 

app/assets/javascripts/discourse/dialects/dialect.js

@@ -379,7 +379,7 @@ Discourse.Dialect = {
       if (!match) { return; }
 
       var lastChance = function() {
-        return !next.some(function(e) { return e.indexOf(args.stop) !== -1; });
+        return !next.some(function(blk) { return blk.match(args.stop); });
       };
 
       // shave off start tag and leading text, if any.
@@ -387,7 +387,8 @@ Discourse.Dialect = {
           leading = block.slice(0, pos),
           trailing = match[2] ? match[2].replace(/^\n*/, "") : "";
       // just give up if there's no stop tag in this or any next block
-      if (block.indexOf(args.stop, pos + args.stop.length) === -1 && lastChance()) { return; }
+      args.stop.lastIndex = block.length - trailing.length;
+      if (!args.stop.exec(block) && lastChance()) { return; }
       if (leading.length > 0) { result.push(['p'].concat(this.processInline(leading))); }
       if (trailing.length > 0) {
         next.unshift(MD.mk_block(trailing, block.trailing,
@@ -405,10 +406,10 @@ Discourse.Dialect = {
           startPos.push(args.start.lastIndex - m[0].length);
           args.start.lastIndex = args.start.lastIndex - (m[2] ? m[2].length : 0);
         }
-        var endPos = [], offset = 0;
-        while ((pos = currentBlock.indexOf(args.stop, offset)) !== -1) {
-          endPos.push(pos);
-          offset += (pos + args.stop.length);
+        args.stop.lastIndex = 0;
+        var endPos = [];
+        while (m = (args.stop).exec(currentBlock)) {
+          endPos.push(args.stop.lastIndex - m[0].length);
         }
 
         // go through the available end tags:
@@ -421,7 +422,8 @@ Discourse.Dialect = {
             // found an end tag, but we must go up a level first.
             ep++; nesting--;
           } else {
-            // found an end tag and we're at the top: done!
+            // found an end tag and we're at the top: done! -- or: start tag and end tag are
+            // identical, (i.e. startPos[sp] == endPos[ep]), so we don't do nesting at all.
             actualEndPos = endPos[ep];
             break blockloop;
           }
@@ -442,8 +444,9 @@ Discourse.Dialect = {
         contentBlocks.push(currentBlock);
       }
 
-      var before = currentBlock.slice(0, actualEndPos).replace(/\n*$/, ""),
-          after = currentBlock.slice(actualEndPos + args.stop.length).replace(/^\n*/, "");
+      var stopLen = currentBlock.match(args.stop)[0].length,
+          before = currentBlock.slice(0, actualEndPos).replace(/\n*$/, ""),
+          after = currentBlock.slice(actualEndPos + stopLen).replace(/^\n*/, "");
       if (before.length > 0) contentBlocks.push(MD.mk_block(before, "", currentBlock.lineNumber));
       if (after.length > 0) next.unshift(MD.mk_block(after, "", currentBlock.lineNumber + countLines(before)));
 

app/assets/javascripts/discourse/dialects/quote_dialect.js

@@ -6,7 +6,7 @@ var esc = Handlebars.Utils.escapeExpression;
 
 Discourse.Dialect.replaceBlock({
   start: new RegExp("\\[quote(=[^\\[\\]]+)?\\]([\\s\\S]*)", "igm"),
-  stop: '[/quote]',
+  stop: /\[\/quote\]/igm,
   emitter: function(blockContents, matches, options) {
 
     var params = {'class': 'quote'},

test/javascripts/lib/markdown-test.js.es6

@@ -365,13 +365,17 @@ test("Code Blocks", function() {
           "<p><pre><code class=\"lang-auto\">hello</code></pre></p>",
           "it doesn't not whitelist all classes");
 
-  cooked("```[quote=\"sam, post:1, topic:9441, full:true\"]This is `<not>` a bug.[/quote]```",
+  cooked("```\n[quote=\"sam, post:1, topic:9441, full:true\"]This is `<not>` a bug.[/quote]\n```",
          "<p><pre><code class=\"lang-auto\">[quote=&quot;sam, post:1, topic:9441, full:true&quot;]This is &#x60;&lt;not&gt;&#x60; a bug.[/quote]</code></pre></p>",
          "it allows code with backticks in it");
 
   cooked("    hello\n<blockquote>test</blockquote>",
          "<pre><code>hello</code></pre>\n\n<blockquote>test</blockquote>",
          "it allows an indented code block to by followed by a `<blockquote>`");
+
+  cooked("``` foo bar ```",
+         "<p><code>foo bar</code></p>",
+         "it tolerates misuse of code block tags as inline code");
 });
 
 test("sanitize", function() {