Terry J. Reedy added the comment:
For future reference, there is an idle specific issue that I do not think has
been mentioned yet. Idle can run in one process, but the default mode now, and
perhaps only mode sometime in the future, is two processes communicating
through a socket. Testing two process communication requires two test
processes, not just one. Tests will pass in one process without testing the
full normal process. For instance:
def fetch_tip(self, expression):
try:
rpcclt = self.editwin.flist.pyshell.interp.rpcclt
except AttributeError:
rpcclt = None
if rpcclt:
return rpcclt.remotecall("exec", "get_the_calltip",
(expression,), {})
else:
return get_argspec(get_entity(expression))
I believe this works because fetch_tip executed in the idle process makes a
remote call that results in fetch_tip being executed in the remote process,
where self has no editwin and hence control falls through to the last line.
A normal, naive test of fetch_tip will 'work' because it will simply fall
through to the last line, which in this case is mostly redundant with separate
tests of get_entity and get_argspec. It seems to me that a full test suite must
at some point test that the actual communication works, and that this cannot
all be done with mocks.
----------
_______________________________________
Python tracker <rep...@bugs.python.org>
<http://bugs.python.org/issue15392>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
