test: Fix memory leaks in cmark.py

Free results of libcmark API function calls.

test/cmark.py

@@ -1,37 +1,63 @@
 #!/usr/bin/env python3
 # -*- coding: utf-8 -*-
 
-from ctypes import CDLL, c_char_p, c_size_t, c_int, c_void_p
+from ctypes import *
 from subprocess import Popen, PIPE
 import platform
 import os
 
+class cmark_mem(Structure):
+    _fields_ = [("calloc", c_void_p),
+                ("realloc", c_void_p),
+                ("free", CFUNCTYPE(None, c_void_p))]
+
 def pipe_through_prog(prog, text):
     p1 = Popen(prog.split(), stdout=PIPE, stdin=PIPE, stderr=PIPE)
     [result, err] = p1.communicate(input=text.encode('utf-8'))
     return [p1.returncode, result.decode('utf-8'), err]
 
 def to_html(lib, text):
+    get_alloc = lib.cmark_get_default_mem_allocator
+    get_alloc.restype = POINTER(cmark_mem)
+    free_func = get_alloc().contents.free
+
     markdown = lib.cmark_markdown_to_html
-    markdown.restype = c_char_p
+    markdown.restype = POINTER(c_char)
     markdown.argtypes = [c_char_p, c_size_t, c_int]
+
     textbytes = text.encode('utf-8')
     textlen = len(textbytes)
     # 1 << 17 == CMARK_OPT_UNSAFE
-    result = markdown(textbytes, textlen, 1 << 17).decode('utf-8')
+    cstring = markdown(textbytes, textlen, 1 << 17)
+    result = string_at(cstring).decode('utf-8')
+    free_func(cstring)
+
     return [0, result, '']
 
 def to_commonmark(lib, text):
-    textbytes = text.encode('utf-8')
-    textlen = len(textbytes)
+    get_alloc = lib.cmark_get_default_mem_allocator
+    get_alloc.restype = POINTER(cmark_mem)
+    free_func = get_alloc().contents.free
+
     parse_document = lib.cmark_parse_document
     parse_document.restype = c_void_p
     parse_document.argtypes = [c_char_p, c_size_t, c_int]
+
     render_commonmark = lib.cmark_render_commonmark
-    render_commonmark.restype = c_char_p
+    render_commonmark.restype = POINTER(c_char)
     render_commonmark.argtypes = [c_void_p, c_int, c_int]
+
+    free_node = lib.cmark_node_free
+    free_node.argtypes = [c_void_p]
+
+    textbytes = text.encode('utf-8')
+    textlen = len(textbytes)
     node = parse_document(textbytes, textlen, 0)
-    result = render_commonmark(node, 0, 0).decode('utf-8')
+    cstring = render_commonmark(node, 0, 0)
+    result = string_at(cstring).decode('utf-8')
+    free_func(cstring)
+    free_node(node)
+
     return [0, result, '']
 
 class CMark: