Issue 34226: cgi.parse_multipart() requires undocumented CONTENT-LENGTH in Python 3.7

In Python 3.7, cgi.parse_multipart() requires "CONTENT-LENGTH" in the second parameter `pdict`. This is not necesary in Python 3.6.

For example, the following code runs with 3.6:

$ python3.6
Python 3.6.6 (default, Jun 27 2018, 13:11:40) 
[GCC 8.1.1 20180531] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import cgi, io
>>> data = b'--heyDavid\r\nContent-Disposition: form-data; name="cfield"\r\n\r\njust a string\r\n\r\n--heyDavid--\r\n'
>>> cgi.parse_multipart(io.BytesIO(data), {"boundary": b"heyDavid"})
{'cfield': [b'just a string\r\n']}

While not on 3.7:

$ python3.7
Python 3.7.0 (default, Jul 15 2018, 10:44:58) 
[GCC 8.1.1 20180531] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import cgi, io
>>> data = b'--heyDavid\r\nContent-Disposition: form-data; name="cfield"\r\n\r\njust a string\r\n\r\n--heyDavid--\r\n'
>>> cgi.parse_multipart(io.BytesIO(data), {"boundary": b"heyDavid"})
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.7/cgi.py", line 220, in parse_multipart
    headers['Content-Length'] = pdict['CONTENT-LENGTH']
KeyError: 'CONTENT-LENGTH'

I looked into the source code of CPython, and found CONTENT-LENGTH in nowhere but cgi.py and test_cgi.py. I guess it has the same meaning as the Content-Length header in HTTP or CONTENT_LENGTH environment variable in CGI? It would be great to document such a backward-incompatible behavior.

Environment: Arch Linux with Python 3.6.6 from [extra] repo and 3.7.0 from [staging] repo.

CC the author and the reviewer of issue29979, where relevant codes are introduced.

working on this on europython

Looks like the change causing this is revision cc3fa20. I would remove the reference to pdict['CONTENT-LENGTH'].

I note this is marked as a 3.7regression and still open. Since the cutoff for the final 3.7 bugfix mode release is in a few days, I'm assuming this means that 3.7 users will have to live with this regression.  If you feel that is a problem, speak up now.

Since there's the easy workaround of just setting pdict["CONTENT-LENGTH"] I personally don't have a problem with this staying the way it is.

But it would be really nice if this were documented somewhere.

I will spend time on this on 12-June or 13-June, 2020.
At the very least, if no changes are required, I will document this for 3.7 and higher versions.

New changeset d8cf3514dd4682419a66f6e834bb384ee34afc95 by roger in branch 'master':
bpo-34226: fix cgi.parse_multipart without content_length (GH-8530)
https://github.com/python/cpython/commit/d8cf3514dd4682419a66f6e834bb384ee34afc95

New changeset aa83935a56d1fd4d72d4de5f0278a240a2d6844d by Miss Islington (bot) in branch '3.7':
[3.7] bpo-34226: fix cgi.parse_multipart without content_length (GH-8530) (GH-20892)
https://github.com/python/cpython/commit/aa83935a56d1fd4d72d4de5f0278a240a2d6844d

New changeset b87453f94fd391e6700eb25dd91de2e56aeeb98f by Miss Islington (bot) in branch '3.9':
bpo-34226: fix cgi.parse_multipart without content_length (GH-8530)
https://github.com/python/cpython/commit/b87453f94fd391e6700eb25dd91de2e56aeeb98f

New changeset c72b7f703eec45c1f4006bf5f65092daedaec46e by Miss Islington (bot) in branch '3.8':
bpo-34226: fix cgi.parse_multipart without content_length (GH-8530)
https://github.com/python/cpython/commit/c72b7f703eec45c1f4006bf5f65092daedaec46e

This is now fixed in all active versions of python. No documentation changes required for this change.