Add escaping to license field

[cdce8p]

Summary of changes

Fix handling of multiline license string by adding rfc822 escaping.

The license is allowed to span over multiple lines, see Metadata spec. Currently, setuptools doesn't autoescape such inputs. The result of which is an PKG-INFO file that isn't parsed correctly by importlib.metadata.

# example setup.cfg
[metadata]
license = This is a pretty 
    long license

Current results

# PKG-INFO
...
Author: ...
License: This is a pretty
long license
Description: This is the package description
        that also spans multiple lines
...

from importlib.metadata import metadata
metadata('custom-pkg')['License']      # -> 'This is a pretty'
metadata('custom-pkg')['Description']  # -> 'long license\nDescription: This ...\n        that ... lines'

With change

# PKG-INFO
...
Author: ...
License: This is a pretty
        long license
Description: This is the package description
        that also spans multiple lines
...

# with change
from importlib.metadata import metadata
metadata('custom-pkg')['License']      # -> 'This is a pretty\n        long license'
metadata('custom-pkg')['Description']  # -> 'This ...\n        that ... lines'

Note that importlib.metadata doesn't remove the escaping. Neither for Description nor License.

Pull Request Checklist

• Changes have tests
• News fragment added in changelog.d/.
  (See documentation for details)

[jaraco]

Note that importlib.metadata doesn't remove the escaping. Neither for Description nor License.

In importlib.metadata 4, it does remove the line continuations.

[EvgenKo423]

The license is allowed to span over multiple lines, see Metadata spec.

1. Core metadata specification refers to RFC 822 for its formatting rules. The reference link at the bottom of this spec describes a convenience representation process called "folding", according to which CRLF followed by any number of spaces is equal to a single space. That is, the example License field does NOT have a multi-line value (there can't be such according to RFC)!
   To overcome this limitation the Description field have to use a | character to be able to preserve line breaks.

2. distutils documentation says that the license key is a

   ‘short string’
         A single line of text, not more than 200 characters.

   which is consistent with the metadata spec, considering the above.

Conclusions (as far as I understood):

1. The license keyword should not be multi-line;
2. The License field of an RFC 822-encoded message can be represented in multiple lines, as well as any other, but should unfold to a single line anyway.

[jaraco]

Yes, I agree. license should not be multiline.