Fix a crashing bug when replacing text in a flow with unicode bodies.

4 files changed, 29 insertions, 7 deletions

diff --git a/libmproxy/flow.py b/libmproxy/flow.py
index 2258868b..6b3d868e 100644
--- a/libmproxy/flow.py
+++ b/libmproxy/flow.py
@@ -249,9 +249,9 @@ class ODict:
         """
         nlst, count = [], 0
         for i in self.lst:
-            k, c = re.subn(pattern, repl, i[0], *args, **kwargs)
+            k, c = utils.safe_subn(pattern, repl, i[0], *args, **kwargs)
             count += c
-            v, c = re.subn(pattern, repl, i[1], *args, **kwargs)
+            v, c = utils.safe_subn(pattern, repl, i[1], *args, **kwargs)
             count += c
             nlst.append([k, v])
         self.lst = nlst
@@ -560,8 +560,8 @@ class Request(HTTPMsg):
             Returns the number of replacements made.
         """
         with decoded(self):
-            self.content, c = re.subn(pattern, repl, self.content, *args, **kwargs)
-        self.path, pc = re.subn(pattern, repl, self.path, *args, **kwargs)
+            self.content, c = utils.safe_subn(pattern, repl, self.content, *args, **kwargs)
+        self.path, pc = utils.safe_subn(pattern, repl, self.path, *args, **kwargs)
         c += pc
         c += self.headers.replace(pattern, repl, *args, **kwargs)
         return c
@@ -740,7 +740,7 @@ class Response(HTTPMsg):
             Returns the number of replacements made.
         """
         with decoded(self):
-            self.content, c = re.subn(pattern, repl, self.content, *args, **kwargs)
+            self.content, c = utils.safe_subn(pattern, repl, self.content, *args, **kwargs)
         c += self.headers.replace(pattern, repl, *args, **kwargs)
         return c
 
@@ -869,7 +869,7 @@ class Error(controller.Msg):
 
             FIXME: Is replace useful on an Error object??
         """
-        self.msg, c = re.subn(pattern, repl, self.msg, *args, **kwargs)
+        self.msg, c = utils.safe_subn(pattern, repl, self.msg, *args, **kwargs)
         return c
 
 
diff --git a/libmproxy/utils.py b/libmproxy/utils.py
index e9c90320..337d4378 100644
--- a/libmproxy/utils.py
+++ b/libmproxy/utils.py
@@ -12,7 +12,7 @@
 #
 # You should have received a copy of the GNU General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
-import os, datetime, urlparse, string, urllib
+import os, datetime, urlparse, string, urllib, re
 import time, functools, cgi
 import json
 
@@ -309,3 +309,15 @@ def parse_size(s):
         return int(s) * mult
     except ValueError:
         raise ValueError("Invalid size specification: %s"%s)
+
+
+def safe_subn(pattern, repl, target, *args, **kwargs):
+    """
+        There are Unicode conversion problems with re.subn. We try to smooth
+        that over by casting the pattern and replacement to strings. We really
+        need a better solution that is aware of the actual content ecoding.
+    """
+    return re.subn(str(pattern), str(repl), target, *args, **kwargs)
+
+
+
diff --git a/test/test_flow.py b/test/test_flow.py
index 627d9cd0..0b7fee4f 100644
--- a/test/test_flow.py
+++ b/test/test_flow.py
@@ -259,6 +259,11 @@ class uFlow(libpry.AutoTree):
         f = flow.Flow(None)
         f.request = tutils.treq()
 
+    def test_replace_unicode(self):
+        f = tutils.tflow_full()
+        f.response.content = "\xc2foo"
+        f.replace("foo", u"bar")
+
     def test_replace(self):
         f = tutils.tflow_full()
         f.request.headers["foo"] = ["foo"]
diff --git a/test/test_utils.py b/test/test_utils.py
index f279ce65..8bff0303 100644
--- a/test/test_utils.py
+++ b/test/test_utils.py
@@ -175,8 +175,13 @@ class u_cleanBin(libpry.AutoTree):
         assert utils.cleanBin("\nne", True) == ".ne"
 
 
+class u_safe_subn(libpry.AutoTree):
+    def test_simple(self):
+        assert utils.safe_subn("foo", u"bar", "\xc2foo")
+
 
 tests = [
+    u_safe_subn(),
     u_cleanBin(),
     u_parse_content_type(),
     uformat_timestamp(),