unittest swallows some lines of the stack trace when raising an AssertionError using the "raise from" syntax inside a TestCase. This marks it harder to pinpoint the source of test failures. It is also confusing to see a stack trace like this because the error doesn't originate where the stack trace says it originates.

To reproduce:

    import unittest

    def foo():
        raise Exception("foo")

    class Test(unittest.TestCase):

        def test_not_okay(self):
            try:
                foo()
            except Exception as exc:
                raise AssertionError("bar") from exc

        def test_okay1(self):
            try:
                foo()
            except Exception as exc:
                raise ValueError("bar") from exc

        def test_okay2(self):
            try:
                foo()
            except Exception as exc:
                raise Exception("bar") from exc

The result (observe how the display for "test_not_okay" differs from the other two):

\======================================================================
ERROR: test_okay1 (error.Test)
\----------------------------------------------------------------------

    Traceback (most recent call last):
      File "/Users/chris/dev/error.py", line 17, in test_okay1
        foo()
      File "/Users/chris/dev/error.py", line 5, in foo
        raise Exception("foo")
    Exception: foo

    The above exception was the direct cause of the following exception:

    Traceback (most recent call last):
      File "/Users/chris/dev/error.py", line 19, in test_okay1
        raise ValueError("bar") from exc
    ValueError: bar

\======================================================================
ERROR: test_okay2 (error.Test)
\----------------------------------------------------------------------

    Traceback (most recent call last):
      File "/Users/chris/dev/error.py", line 23, in test_okay2
        foo()
      File "/Users/chris/dev/error.py", line 5, in foo
        raise Exception("foo")
    Exception: foo

    The above exception was the direct cause of the following exception:

    Traceback (most recent call last):
      File "/Users/chris/dev/error.py", line 25, in test_okay2
        raise Exception("bar") from exc
    Exception: bar

\======================================================================
FAIL: test_not_okay (error.Test)
\----------------------------------------------------------------------

    Traceback (most recent call last):
      File "/Users/chris/dev/error.py", line 11, in test_not_okay
        foo()
    Exception: foo

    The above exception was the direct cause of the following exception:

    Traceback (most recent call last):
      File "/Users/chris/dev/error.py", line 13, in test_not_okay
        raise AssertionError("bar") from exc
    AssertionError: bar

\----------------------------------------------------------------------
Ran 3 tests in 0.001s

FAILED (failures=1, errors=2)

I guess this isn't limited just to the "raise from" syntax. It also occurs if "from exc" is removed from the example above.

Hmm, this is a little surprising, but - why are you raising AssertionError like that - thats what assertRaises is for.

This is simply a reduced test case to illustrate the issue more clearly. There was more to it in how I was using it in practice.

I'm fairly sure its eating the stack frames because the calling frames are annotated __unittest__ - its technically a feature :/.

Yes, it is... ish.

The frame skipping code occurs when we serialise exceptions, and we pass a limit in. The limit is calculated on the main exception only. If the cause has a longer exception than the limit we calculated, you'd see this behaviour.

Probably need to make it possible to do per-exception processing of limit: I think via a callback or similar mechanism in traceback (because the cause might have thrown from some code that is also marked __unittest, so if we're honouring that, we should honour it within each exception.

I've run into this bug too. Took a while to track down the cause :)

I think this is the same as bpo-42247, where I've recently attached a patch that removes __unittest frames from the traceback rather than count them and then set a limit for printing.

I could add to that patch what Robert suggests here - to filter the __cause__ traceback as well.

New changeset 88b7d86 by Irit Katriel in branch 'main':
bpo-24959: fix unittest.assertRaises bug where traceback entries are dropped from chained exceptions (GH-23688)
88b7d86

New changeset 26fa25a by Miss Islington (bot) in branch '3.10':
bpo-24959: fix unittest.assertRaises bug where traceback entries are dropped from chained exceptions (GH-23688)
26fa25a

New changeset f3ea249 by Irit Katriel in branch '3.9':
bpo-24959: fix unittest.assertRaises bug where traceback entries are dropped from chained exceptions (GH-23688) (GH-31776)
f3ea249