endless recursion checking the bool_value of a condition with value bool()

[ltcmelo]

I'm facing an endless recursion (apparently, from the trace, not the same as this currently open issue). Consider this snippet:

v = bool() or 1

When I infer the value of v, the execution gets to a point where it bounces back and forth between def _infer_method_result_truth and def bool_value. Eventually, the maximum recursion limit is reached.

I believe that the root cause of the problem is here: due to bool(), we have the inferred object as being the "same" as the instance object (an instance of class bool). I suppose that a check* must be added there, for this equality.

*I could create a PR, but it seems that Instance has no __eq__ method; I could still check with if (str(inferred) == str(instance): return util.Uninferable but I'm not sure it'd be corred/desired.

NOTE: The same issue happens in slight variations, e.g., v = w if hasattr(o, 'x') else 1

[hippo91]

@ltcmelo thanks for your report.
I'm am currently unable to reproduce it.
Can you give a snippet of the problematic code and the version of python and astroid you are using?

[ltcmelo]

Hi @hippo91, it seems that:

• not always the recursion is infinite, there appears to exists a co-relation with the files/project being processed; I couldn't spot what is breaking the recursion, though — I suspect that, in certain cases, the context object passed to infer will aggregate enough info, but those cases don't seem to be guaranteed.
• even with a limited recursion, it's still possible to observe (what I believe is) the bogus behavior I mentioned.

A list of steps that you could try in order to validate my suspicion. Perhaps, you could try reducing your sys.setrecursionlimit too…

1. Right after the statement I pointed to earlier, add the following:

print(f'\n[[inferred:{inferred}]] [[value:{value}]] [[context:{context}]]')

2. Create a regular AST visitor, load a module with that snippet I originally suggested, and, upon visit_name, call node.infer on object corresponding to v.

I expect that you see tens of calls jumping back and forth between those 2 methods, repeatedly trying to infer the same bool instance. This is how my prints look like.

Python version: Python 3.8.6 (CPython on darwin)
astroid: 2.4.2

[hippo91]

@ltcmelo thanks for your update. Sadly without a clean snippet that reproduce the problem, i'm afraid there is not so much we can do here...