Trial's grouping of failed tests with stack traces is confusing and ambiguous

	@jml reported
Trac ID	trac#7935
Type	defect
Created	2015-06-11 11:15:42Z

I recently ran a large test suite and at the end of the run got these failures.

flocker.control.test.test_httpapi.MemoryTestsUpdatePrimaryDataset.test_unknown_primary_node
flocker.control.test.test_httpapi.RealTestsUpdatePrimaryDataset.test_unknown_primary_node
===============================================================================
[ERROR]
Traceback (most recent call last):
  File "/vagrant/.env/lib/python2.7/site-packages/twisted/internet/defer.py", line 140, in maybeDeferred
    result = f(*args, **kw)
… traceback elided …
  File "/vagrant/.env/lib/python2.7/site-packages/twisted/trial/_synctest.py", line 447, in assertEqual
    % (msg, pformat(first), pformat(second)))
twisted.trial.unittest.FailTest: not equal:
a = {'line': 'mkdir: cannot create directory \xe2\x80\x98/.virtualenvs\xe2\x80\x99: Permission denied'}
b = {'line': 'test_ssh_conch:test_run_logs_stderr'}


flocker.provision.test.test_ssh_conch.Tests.test_run_logs_stderr
===============================================================================
[ERROR]
Traceback (most recent call last):
  File "/vagrant/.env/lib/python2.7/site-packages/twisted/internet/defer.py", line 140, in maybeDeferred
    result = f(*args, **kw)
… traceback elided …
  File "/vagrant/.env/lib/python2.7/site-packages/twisted/trial/_synctest.py", line 447, in assertEqual
    % (msg, pformat(first), pformat(second)))
twisted.trial.unittest.FailTest: not equal:
a = {'line': 'mkdir: cannot create directory \xe2\x80\x98/.virtualenvs\xe2\x80\x99: Permission denied'}
b = {'line': 'test_ssh_conch:test_run_logs_stdout'}


flocker.provision.test.test_ssh_conch.Tests.test_run_logs_stdout
-------------------------------------------------------------------------------
Ran 208 tests in 7.910s

I assumed that the test names at the top (flocker.control.test.test_httpapi.MemoryTestsUpdatePrimaryDataset.test_unknown_primary_node and so forth) were the cause of the errors, perhaps because the long double underline reminded me of the ReST heading syntax, and because I assume that the traceback is one of the details about the failing test.

This might be subjective. Perhaps the best way of fixing it is to provide an alternative formatter.

trac-id__7935 7935
type__defect defect
reporter__jml jml
priority__normal normal
milestone__ 
branch__ 
branch_author__ 
status__new new
resolution__None None
component__trial trial
keywords__ 
time__1434021342040445 1434021342040445
changetime__1434021342113465 1434021342113465
version__None None
owner__None None
cc__jml

Start by reproducing the large test-suite output shown in the report and read how Twisted Trial groups failed test names, tracebacks, and error details. Define a formatter or presentation change that makes each failure and its traceback unambiguous, then verify the output against the reported examples.