Test also markdown serialization

As two different markdown strings can be semantically the same,
we can test if the resulting rendered HTML are syntactically equal.

Currently about hundred test cases fail because of a known limitation of
`prosemirror-markdown` not able to serialize nested marks correctly, see
ProseMirror/prosemirror-markdown#82

Signed-off-by: Ferdinand Thiessen <rpm@fthiessen.de>

src/tests/markdown.spec.js

@@ -25,9 +25,11 @@ const markdownThroughEditorHtml = (html) => {
 describe('Commonmark', () => {
 	const skippedMarkdownTests = [
 		// contain HTML
-		21, 31, 201, 344, 474, 475, 476, 490, 493, 523, 535, 642, 643,
+		21, 31, 344, 474, 475, 476, 642, 643,
 		// contain comments
 		309, 308,
+		// < > are escaped, because HTML is disabled for markdown-it
+		201, 490, 493, 523, 535
 	];
 
 	const normalize = (str) => {
@@ -52,6 +54,18 @@ describe('Commonmark', () => {
 			// Ignore special markup for untouched markdown
 			expect(normalize(rendered)).toBe(expected)
 		})
+
+		test('commonmark serialization ' + entry.example, () => {
+			const expected = markdownit.render(entry.markdown)
+			const serialized = markdownThroughEditorHtml(expected)
+
+			try {
+				expect(markdownit.render(serialized)).toBe(expected)
+			} catch(e) {
+				// This is just for debugging, so jest shows also the difference within the two markdown source codes
+				expect(markdownit.render(serialized) + '\n\n' + serialized).toBe(expected + '\n\n' + entry.markdown)
+			}
+		})
 	})
 })