Not sure I can really call this a bug, however there is a behavioral change between 2.7 and at least 3.6 and 3.7 (probably earlier versions of the 3.x series as well). There is no spec for .netrc files that I can find, certainly nothing which mentions comment or blank lines. Still, Python's netrc file parser seems happy with both.

However, in 3.x a blank line followed immediately by a comment line containing actual comment text causes the parser to raise a parse error. I've attached two netrc files, netrc-comment-blank, and netrc-blank-comment, identical save for the ordering of a blank line and a comment line. Here's what a 2.7.14 session looks like:

Python 2.7.14 |Anaconda, Inc.| (default, Mar 27 2018, 17:29:31) 
[GCC 7.2.0] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import netrc
>>> rc = netrc.netrc(file="/home/skip/tmp/netrc-comment-blank")
>>> rc = netrc.netrc(file="/home/skip/tmp/netrc-blank-comment")

Here's 3.7.0:

Python 3.7.0 (default, Jun 28 2018, 13:15:42) 
[GCC 7.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import netrc
>>> rc = netrc.netrc(file="/home/skip/tmp/netrc-comment-blank")
>>> rc = netrc.netrc(file="/home/skip/tmp/netrc-blank-comment")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/skip/miniconda3/envs/python3/lib/python3.7/netrc.py", line 30, in __init__
    self._parse(file, fp, default_netrc)
  File "/home/skip/miniconda3/envs/python3/lib/python3.7/netrc.py", line 63, in _parse
    "bad toplevel token %r" % tt, file, lexer.lineno)
netrc.NetrcParseError: bad toplevel token 'Comment' (/home/skip/tmp/netrc-blank-comment, line 2)

This seems to have been introduced with commit : 1df0f214a9fdb4dde7506576b144cf6a7fd01b65 before which it was working fine.

Testing shows the error from 1df0f214a9fdb4dde7506576b144cf6a7fd01b65 and above. I have checked out only versions of Lib/netrc.py and I think this doesn't require a rebuild of Python.

➜  cpython git:(master) $ git log --oneline --format="%h" Lib/netrc.py > commits.txt # Get all commits to test the changes
➜  cpython git:(master) ✗ ./python netrc_foo_tester.py

netrc_foo_tester.py is attached with the message which basically does the below using Python for every commit. netrc_foo.py contains the actual test of parsing two files.

➜  cpython git:(master) ✗ ./python
Python 3.8.0a0 (heads/master:35c0809, Jul 16 2018, 10:29:23)
[GCC 5.4.0 20160609] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>>

➜ cpython git:(master) ✗ cat netrc_foo.py
import netrc

rc = netrc.netrc(file="netrc-comment-blank")
rc = netrc.netrc(file="netrc-blank-comment")

➜  cpython git:(master) ✗ git reset --quiet HEAD . && git checkout . && git checkout be19ed7 Lib/netrc.py && ./python netrc_foo.py
netrc_foo.py:2: ResourceWarning: unclosed file <_io.TextIOWrapper name='netrc-comment-blank' mode='r' encoding='UTF-8'>
  rc = netrc.netrc(file="netrc-comment-blank")
netrc_foo.py:3: ResourceWarning: unclosed file <_io.TextIOWrapper name='netrc-blank-comment' mode='r' encoding='UTF-8'>
  rc = netrc.netrc(file="netrc-blank-comment")
➜  cpython git:(master) ✗ git reset --quiet HEAD . && git checkout . && git checkout c12a813 Lib/netrc.py && ./python netrc_foo.py
➜  cpython git:(master) ✗ git reset --quiet HEAD . && git checkout . && git checkout 78a1a15 Lib/netrc.py && ./python netrc_foo.py
➜  cpython git:(master) ✗ git reset --quiet HEAD . && git checkout . && git checkout 1df0f21 Lib/netrc.py && ./python netrc_foo.py
Traceback (most recent call last):
  File "netrc_foo.py", line 3, in <module>
    rc = netrc.netrc(file="netrc-blank-comment")
  File "/home/cpython/Lib/netrc.py", line 32, in __init__
    self._parse(file, fp)
  File "/home/cpython/Lib/netrc.py", line 65, in _parse
    "bad toplevel token %r" % tt, file, lexer.lineno)
netrc.NetrcParseError: bad toplevel token 'Comment' (netrc-blank-comment, line 2)

I hope the above approach is correct.

Thanks

I took a shot at this, but closed my PR when I found that it introduced a different regression.

After playing with it for a bit, I think that this commit [1] is a bit problematic because it uses lexer.instream.readline(), which can make the line number in lexer.lineno incorrect.

I may take another look if I have time, but someone else is free to tackle this.

[1] https://github.com/python/cpython/commit/1df0f214a9fdb4dde7506576b144cf6a7fd01b65#diff-e6896f6d68ca3fd8094a933533f8b2ed

I realized that the Python 2 solution was adapt-able after all; I've re-submitted as GitHub PR 8360.

https://github.com/python/cpython/pull/8360

messages: + msg321810