2.7.3 regression ignores all disabled warnings

[fmigneault]

Steps to reproduce

Moving from 2.7.2 to 2.7.3, my repo goes from no error to hundreds of problems.
I didn't go through all of them, but many (if not all) seem to be related to ignored errors.
My config: https://github.com/Ouranosinc/Magpie/blob/master/.pylintrc

Current behavior

All these errors are flagged, when they are not just a patch version before.

Messages
--------

+-------------------------------+------------+
|message id                     |occurrences |
+===============================+============+
|unnecessary-lambda             |54          |
+-------------------------------+------------+
|ungrouped-imports              |46          |
+-------------------------------+------------+
|no-self-use                    |30          |
+-------------------------------+------------+
|super-with-arguments           |27          |
+-------------------------------+------------+
|too-many-locals                |25          |
+-------------------------------+------------+
|broad-except                   |21          |
+-------------------------------+------------+
|import-outside-toplevel        |17          |
+-------------------------------+------------+
|cell-var-from-loop             |17          |
+-------------------------------+------------+
|useless-object-inheritance     |14          |
+-------------------------------+------------+
|unused-argument                |12          |
+-------------------------------+------------+
|comparison-with-callable       |11          |
+-------------------------------+------------+
|invalid-envvar-default         |10          |
+-------------------------------+------------+
|wrong-import-order             |9           |
+-------------------------------+------------+
|raise-missing-from             |8           |
+-------------------------------+------------+
|inconsistent-return-statements |6           |
+-------------------------------+------------+
|use-a-generator                |5           |
+-------------------------------+------------+
|too-many-public-methods        |5           |
+-------------------------------+------------+
|len-as-condition               |4           |
+-------------------------------+------------+
|useless-super-delegation       |2           |
+-------------------------------+------------+
|unnecessary-comprehension      |2           |
+-------------------------------+------------+
|too-many-statements            |2           |
+-------------------------------+------------+
|too-many-branches              |2           |
+-------------------------------+------------+
|try-except-raise               |1           |
+-------------------------------+------------+
|import-error                   |1           |
+-------------------------------+------------+

-------------------------------------------------------------------
Your code has been rated at 9.77/10 (previous run: 10.00/10, -0.23)

Expected behavior

No more warning/errors than before, as code didn't change at all.

pylint --version output

Result of pylint --version output:

pylint 2.7.3
astroid 2.5.2
Python 3.7.9 (default, Aug 31 2020, 12:42:55) 
[GCC 7.3.0]

[cdce8p]

Please try reformatting your .pylintrc file:

disable=
    missing-docstring,  # C0111
    too-many-lines,  # C0302

The , after each entry is required (except for the last one). I noticed that it was missing after: no-self-use.

[cdce8p]

@Pierre-Sassoulas f501b87 seems to cause this behavior. My suggestion works, so I'm not sure how to best deal with it.

I used this for testing. Python file a.py, pylintrc: customrc

# pylint: disable=unused-import,import-error,missing-docstring
import const

import typing

[MESSAGES CONTROL]

disable=
    C0111,missing-docstring,
    C0302,too-many-lines,
    C0330,bad-continuation,
    C0411,wrong-import-order,

pylint --rcfile customrc a.py

After the commit, it outputs a wrong-import-order error.

[calvincheng8]

I have the following config for test.rc,

[MESSAGES CONTROL]
disable=C0111,C0326,W0703

Pylint generates a warning W0703 that should have been ignored on the following file test.py,

try:
    f = open('test')
except Exception:
    pass

$ pylint --rcfile `test.rc` test.py
************* Module test
test.py:3:7: W0703: Catching too general exception Exception (broad-except)

------------------------------------------------------------------
Your code has been rated at 7.50/10 (previous run: 7.50/10, +0.00)

Removing C0326 from the disable list and the disable works as expected.

Pylint probably is failing silently when it encounters C0326.

Running on command line confirms this,

$ pylint -d C0326 test.py

throws a KeyError: 'C0326'.

[Pierre-Sassoulas]

My bad, I did not add a functional test for disabling with pylintrc when I refactored this. I'm going to release 2.7.4 and 3.0.0a1 immediately as it is very problematic.

[fmigneault]

@cdce8p
Thanks for spotting that!

Is there any chance (or how if already possible) this kind of thing can be detected and flagged as warning?
I imagine the disable field gets parsed and split into individual components, and could easily indicate that some definition is unknown by printing its name. Since it wouldn't be split, it would print no-self-useR0205 in my case and I could quickly figure out that a parsing error is the cause.

[Pierre-Sassoulas]

I'm working on it right now @fmigneault :)