Issue 27033: Change the decode_data default in smtpd to False

Proposed patch changes the default value of the decode_data parameter for smtpd.SMTPChannel and smtpd.SMTPServer constructors. The default value True was deprecated since adding this parameter in issue19662.

New changeset a31b9f353346 by Serhiy Storchaka in branch 'default':
Issue #27033: The default value of the decode_data parameter for
https://hg.python.org/cpython/rev/a31b9f353346

Um, I changed to to commit review because I was planning to review it.  I don't expect to find any problems, but I'm surprised you committed it.

There does need to be a whatsnew entry, and specifically something in the porting section, since this is a behavior change.

aiosmtpd already defaults decode_data to False.  We should consider pulling this into 3.6 (not just for this reason of course).

http://aiosmtpd.readthedocs.io/en/latest/

Oh, sorry, I understood this as a sign that you already made a review and don't have comments. I was surprised that you made this silently.

Here is a patch that adds an entry in the porting section of What's New. Please also check changes in the smtpd documentation. I reformulated some sentences.

No problem.  I use 'commit review' as an indication that a committer should do a final review...before that, triage or general community people can be doing the reviews, then triage can move it to commit review to request the final review before commit.  What this means in practical terms is that if I do a search for 'email' issues in 'commit review' stage, I find all the issues I should be reviewing.

I'll try to do the review soon.

Thank you for your review David. Updated patch addresses your doc comments.

As for bool calls, I think there are reasons to do this.

Addressed other David's comment.

What reasons?  It's not a big deal, but like I say it isn't our normal style.

Oh, I forgot to say that the revisions look good to me.

Since these arguments are saved for multiple testing, it is worth to ensure that every test returns consistent result. For public attribute there is additional benefit, it looks as boolean.

New changeset 9abde519cc1e by Serhiy Storchaka in branch 'default':
Improved docs for issue27033. Based on comments by R. David Murray.
https://hg.python.org/cpython/rev/9abde519cc1e

Python uses duck typing, though.  Maybe someone is depending on those public attributes not getting coerced to boolean.  Probably not, granted, since they are newish, but a python programmer might expect the value to get passed through.  I'm of two minds about the advisability of that, but given that someone *could* be depending on it, I think it would be better to not do the bool call, since there is no bug fixed by doing the bool call, making it an unnecessary change in behavior.

New changeset 71813a05e488 by Serhiy Storchaka in branch 'default':
Issue #27033: Removed unnecessary the bool calls.
https://hg.python.org/cpython/rev/71813a05e488

Can this issue be closed?

Yes.  Thanks, Serhiy.

stage: commit review -> resolved

messages: + msg266626

messages: + msg266625

messages: + msg265716

stage: resolved -> commit review