>>> s = "x" * (2**29)
>>> case.assertEqual(s + "a", s + "b")
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
 File "/home/antoine/cpython/default/Lib/unittest/case.py", line 643,
in assertEqual assertion_func(first, second, msg=msg)
 File "/home/antoine/cpython/default/Lib/unittest/case.py", line 984,
in assertMultiLineEqual secondlines = [second + '\n']
MemoryError

assertEqual delegates to assertMultilineEqual for comparing text which uses difflib for comparisons. This has performance issues (as well as memory issues) for very large inputs, so should fallback to a simple comparison (or simpler diff generation technique) for very large inputs.

Attached patch adds a _diffThreshold attribute of 2**16 and uses _baseAssertEqual whenever one of the two string is longer than 2**16 chars.

Rather than hardwiring `self.addCleanup(lambda: setattr(self, '_diffThreshold', 2**16))`, you should retrieve the previous value.

New changeset 8dbf661c0a63 by Ezio Melotti in branch '2.7':
#11763: don't use difflib in TestCase.assertMultiLineEqual if the strings are too long.
http://hg.python.org/cpython/rev/8dbf661c0a63

New changeset 04e64f77c6c7 by Ezio Melotti in branch '3.1':
#11763: don't use difflib in TestCase.assertMultiLineEqual if the strings are too long.
http://hg.python.org/cpython/rev/04e64f77c6c7

New changeset b316019638df by Ezio Melotti in branch '3.2':
#11763: merge with 3.1.
http://hg.python.org/cpython/rev/b316019638df

New changeset df154c872b0c by Ezio Melotti in branch 'default':
#11763: merge with 3.2.
http://hg.python.org/cpython/rev/df154c872b0c

I committed a slightly modified version of the patch, so the issue should be fixed now.

There are two related problems though:
1) difflib is used in other places, so those should probably be checked too;
2) _baseAssertEqual should check if the len of the msg (or the sum of the lengths of the two args) is greater than maxDiff and use safe_repr(arg, short=True) to avoid long diffs.

Here is a small patch addressing the second point of the last message, but the tests are still missing.

The basic idea of the patch is good, but instead of introducing _MAX_LENGTH, maxDiff should be reused.

Sorry, ignore that. I see that the patch already passes maxDiff to truncate_str.

I left comments on Rietveld. Tests needed.

There are some tests in the issue11763-2.diff patch.

Attached an updated patch that addresses Serhiy comments.  The tests in the previous patch have been committed already, so this might need new tests.

The first problem I mentioned in msg134530 has been reported in #19217.

The second problem I mentioned in msg134530 got fixed in #18996, making the patches on this issue obsolete.
Since the other problem is now tracked in #19217, I'm going to close this.