bpo-18019 noted the following crash in earlier 2.7:

>>> d={}
>>> d[42]=d.viewvalues()
>>> d
<segmentation fault>

This issue has been fixed; the behaviour now is that a RuntimeError is
produced for a recursive dictionary view:

>>> d={}
>>> d[42]=d.viewvalues()
>>> d # (output line-broken:)
{42: Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RuntimeError: maximum recursion depth exceeded
  while getting the repr of a list

Before finding this, though, I'd investigated and made a patch which
produces a similar "..." output to a recursive dictionary. Reworking
against current 2.7, the behaviour would be:

>>> x={}
>>> x[42]=x
>>> x # existing behaviour for dictionaries:
{42: {...}}

>>> d={}
>>> d[42]=d.viewvalues()
>>> d # new behaviour:
{42: dict_values([...])}
>>> d[43]=d.viewitems()
>>> d # (output line-broken:)
{42: dict_values([..., dict_items([(42, ...), (43, ...)])]),
 43: dict_items([(42, dict_values([..., ...])), (43, ...)])}

Attached is the patch, against current 2.7 branch. If there is interest
in applying this, I will create a proper patch (changelog entry, fix to
Lib/test/test_dictviews.py, etc.).

On python-dev, Gregory Smith noted:

Given that the RuntimeError fix has been released, your proposed
... behavior is arguably a new feature so I'd only expect this to
make sense for consideration in 3.4, not 2.7.  (if accepted at all)

[http://mail.python.org/pipermail/python-dev/2013-July/127489.html]

so I have marked for consideration for versions 2.7 and 3.4.

Fixing 3.3 is more likely that 3.4. I can view RuntimeError as a bug less obnoxious than crash, but others may differ. Ultimately, the release managers can decide. This is definitely appropriate for 3.4, so please add tests. If the patch is for 2.7, please do one for 3.3 or 3.4.

For 3.x, d.keys() is not an issue, but d.values() is.

>>> d = {}
>>> d[1] = d.keys()
>>> d
{1: dict_keys([1])}
>>> d[2] = d.values()
>>> d
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RuntimeError: maximum recursion depth exceeded while getting the repr of a list

As I wrote on the list -- IMHO it's still a bug (even though not so painful as segfault) that should also be fixed in 2.7 and 3.2/3.3. In other cases (such as d={}; d[42]=d; repr(d)) Python does its best to avoid an error -- why in this case (d={}; d[42]=d.<Py2.x:view>values(); repr(d)) should it raise an exception? IMHO it's an obvious oversight in implementation, not a feature that anybody would expect.

New patch including tests attached, against 3.3. Terry Reedy's example
above now gives

    >>> d = {}
    >>> d[1] = d.keys()
    >>> d[2] = d.values()
    >>> d
    {1: dict_keys([1, 2]), 2: dict_values([dict_keys([1, 2]), ...])}

which I think is preferable.

Summary of patch:

dictobject.c:

dictview_repr() now uses a Py_ReprEnter() / Py_ReprLeave() pair to
check for recursion, and produces "..." if so.  I think I've got the
behaviour correct in the case that PySequence_List() fails, but
couldn't find a way to trigger this in testing.

test_dictviews.py:

test_recursive_repr() checks for the correct string (containing
"...") rather than a RuntimeError.

test_deeply_nested_repr() is a new function, aiming to replace the
original test_recursive_repr().  It checks that a RuntimeError is
raised in the case of a non-recursive but deeply nested structure.

Under --with-pydebug,

./python -m test.regrtest -R20:30 test_dictviews

passes.

non-error-recursive-dictview-3.3.patch: small nit, initialize result to NULL, instead of setting it to NULL only in case of error. It makes the code more readable and it is the common coding style to handle errors.

New patch, fixing nit noted in msg193624:

non-error-recursive-dictview-3.3-1.patch

Otherwise same as previous

non-error-recursive-dictview-3.3.patch

I'm in favor of treating this a bugfix. It is a bug for repr to raise a recursion error, IMO.

+1 for backporting the fix.

I'll attach fresh patches, one against latest 2.7 and one against
latest 3.3 (unchanged apart from headers from previous patch).

Both branches pass

./python -m test.regrtest -R20:30 test_dictviews

Is anything further needed from me before this can be reviewed?

Please visit
http://www.python.org/psf/contrib/
http://www.python.org/psf/contrib/contrib-form/
and submit a Contributor Agreement. This process is complete when '*' appears after your name here, as with mine and others.

Ben, would you be interested in making a Github pull request against the current master for your patch? It appears there was interest in moving forward with this and it is still occurring on 3.7, although it now raises a RecursionError.

>>> d = {}
>>> d[1] = d.keys()
>>> d
{1: dict_keys([1])}
>>> d[2] = d.values()
>>> d
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RecursionError: maximum recursion depth exceeded while getting the repr of an object

See also bpo-32137.

PR4823 created as per msg308086.

It looks like Serhiy approved the PR on 12/13. Is it ready to merge? Thanks!

Not ready to merge.

1. We need the Appveyer test to pass or special action from the gateway gods.
2. After a month, the PR needed to be updated and retested anyway. I pulled, updated, pushed the update (which will trigger Travis and AV), and tested on my Windows machine. The two new tests fail. See PR.
3. Someone has to sell assign and merge when ready.

Hold the failure comment. I forgot that I have to recompile when patch changes .c files.

OK, test_dictviews and test_ordered_dict now pass for me, and Appveyor already ran entire suite and says OK to all.

New changeset d7773d9 by Senthil Kumaran (bennorth) in branch 'master':
bpo-18533: Avoid RecursionError from repr() of recursive dictview (bpo-4823)
d7773d9

I merged this one. All the requirements were met.

@terry / @ Cheryl - could you please review if I did it right? Should this be backported to other Python versions?

Thank you for checking and merging. In the future, we will be reminded to change '#' to 'GH-' in the title. I only found out a couple of days ago that this is desired.

David, in msg193651, and Raymond, in msg193779, already suggested backporting. I added the 'needs backport' tags.

The auto backport for 3.6 worked. You might want to checked the diff as well as test results.

As I expected from the fact that Ben once did a separate 2.7 version, the auto backport for 2.7 failed. Ben, since you know what code changes are needed, can you prepare a backport (cherry pick) PR?

Terry in msg310785:

As I expected from the fact that Ben once did a separate 2.7
version, the auto backport for 2.7 failed. Ben, since you know what
code changes are needed, can you prepare a backport (cherry pick)
PR?

Yes, that's right, there were some differences between the approaches
in Py2 and Py3. I've created PR5357 for the 2.7 backport.

New changeset fbf7aac by Miss Islington (bot) in branch '3.6':
bpo-18533: Avoid RecursionError from repr() of recursive dictview (GH-4823)
fbf7aac