pylint complains about: "E1101: Instance of '' has no 'Object' member (no-member)" for boto3 resources

[mludvig]

Somehow pylint 2.3 and newer started complaining about a simple boto3 resources. It can be easily reproduced in a clean docker container.

1. Start the container and install the newest boto3 and pylint.

   $ docker run --rm -it python:3.7 bash
   root@0ebc7eb0b8d9:/# pip install pylint boto3
   [...]
   Successfully installed astroid-2.3.3 boto3-1.10.20 botocore-1.13.20 docutils-0.15.2 isort-4.3.21 jmespath-0.9.4 lazy-object-proxy-1.4.3 mccabe-0.6.1 pylint-2.4.4 python-dateutil-2.8.0 s3transfer-0.2.1 six-1.13.0 typed-ast-1.4.0 urllib3-1.25.7 wrapt-1.11.2
   

2. Create a very simple test file:

   root@0ebc7eb0b8d9:/# cat > test-pylint-boto3.py 
   import boto3
   
   sqs = boto3.resource('sqs')
   sqs_queue = sqs.Queue(url='http://...')         # <<< pylint doesn't like this
   
   s3 = boto3.resource('s3')
   s3_obj = s3.Object(bucket_name='boto3', key='test.py')            # <<< and this
   

3. Now run pylint:

   root@0ebc7eb0b8d9:/# pylint -E test-pylint-boto3.py 
   ************* Module test-pylint-boto3
   /test-pylint-boto3.py:4:12: E1101: Instance of '' has no 'Queue' member (no-member)
   /test-pylint-boto3.py:7:9: E1101: Instance of '' has no 'Object' member (no-member)
   

4. The latest one that worked was pylint == 2.2.3. Downgrade to that version and verify clean run:

   root@0ebc7eb0b8d9:/# pip install 'pylint<2.3'
   Collecting pylint<2.3
   ...
   Successfully installed astroid-2.1.0 pylint-2.2.3
   
   root@0ebc7eb0b8d9:/# pylint -E test-pylint-boto3.py 
   root@0ebc7eb0b8d9:/# 
   

TL;DR
With pylint==2.3.0 something changed and pylint can no longer be used with boto3. It doesn't depend on boto3 versions, I tried a few major ones back and the new pylint keeps throwing this error regardless of the boto3 version.

Can we get boto3 support back please?

[PCManticore]

Thanks for the report. This is a duplicate of #3134. Also it's not like we "took" boto3 support from you intentionally. There probably is a bug somewhere in the inference which prevents properly inferring boto3 types, but we did not have time yet to further investigate the cause of that.