StrEnum subclasses cannot be created

[ethanfurman]

BPO	34909
Nosy	@ned-deily, @ethanfurman, @matrixise
PRs	bpo-34909: keep searching mixins until base class is found #9737 [3.7] bpo-34909: keep searching mixins until base class is found (GH-9737) #9738 bpo-34909: NEWS entry #9995

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/ethanfurman'
closed_at = <Date 2018-10-20.17:08:50.098>
created_at = <Date 2018-10-05.20:12:48.209>
labels = ['type-bug', '3.7']
title = 'StrEnum subclasses cannot be created'
updated_at = <Date 2018-10-20.17:16:13.436>
user = 'https://github.com/ethanfurman'

bugs.python.org fields:

activity = <Date 2018-10-20.17:16:13.436>
actor = 'ethan.furman'
assignee = 'ethan.furman'
closed = True
closed_date = <Date 2018-10-20.17:08:50.098>
closer = 'ned.deily'
components = []
creation = <Date 2018-10-05.20:12:48.209>
creator = 'ethan.furman'
dependencies = []
files = []
hgrepos = []
issue_num = 34909
keywords = ['patch', '3.7regression']
message_count = 22.0
messages = ['327182', '327207', '327209', '327210', '327215', '327216', '327228', '327234', '327245', '327254', '327624', '327773', '327794', '327866', '327930', '328123', '328140', '328144', '328153', '328166', '328169', '328172']
nosy_count = 3.0
nosy_names = ['ned.deily', 'ethan.furman', 'matrixise']
pr_nums = ['9737', '9738', '9995']
priority = None
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue34909'
versions = ['Python 3.7']

[ethanfurman]

from enum import Enum, unique

class StrEnum(str, Enum):
    def __new__(cls, *args, **kwargs):
        for a in args:
            if not isinstance(a, str):
                raise TypeError("Enumeration '%s' (%s) is not"
                                " a string" % (a, type(a).__name__))
        return super(StrEnum, cls).__new__(cls, *args, **kwargs)

@unique
class Decision(StrEnum):
    """Decision results/strategy enumeration."""
    REVERT = "REVERT"
    REVERT_ALL = "REVERT_ALL"
    RETRY = "RETRY"

---

Traceback (most recent call last):
  File "test", line 14, in <module>
    class Decision(StrEnum):
  File ".../cpython/Lib/enum.py", line 222, in __new__
    enum_member._value_ = member_type(*args)
  File ".../cpython/Lib/enum.py", line 309, in __call__
    return cls.__new__(cls, value)
  File ".../cpython/Lib/enum.py", line 545, in __new__
    return cls._missing_(value)
  File ".../cpython/Lib/enum.py", line 558, in _missing_
    raise ValueError("%r is not a valid %s" % (value, cls.__name__))
ValueError: 'REVERT' is not a valid StrEnum

[ned-deily]

Ethan, do you an ETA of a fix for this? 3.7.1 final is scheduled for fewer than 24 hours from now but if it's truly a release blocker I will hold for a fix and probably do a rc2 instead.

[ethanfurman]

It really is. I got a message from Michal Arbet from OpenStack about it.

I just finished the fix and am currently running the tests locally. Is it okay to have the PR directly against 3.7 instead of doing 3.8 first and backporting?

[ned-deily]

I just finished the fix and am currently running the tests locally.

Great. Thanks!

Is it okay to have the PR directly against 3.7 instead of doing 3.8 first and backporting?

If you want but it's probably easier to do it the normal way and let the bot do the backport, assuming master and 3.7 are still similar in that area.

[matrixise]

Hi Ethan,

the issue appears with your patch

0c076ca is the first bad commit
commit 0c076ca
Author: Ethan Furman <ethan@stoneleaf.us>
Date: Fri Sep 21 22:26:32 2018 -0700

[3.7] bpo-29577: Enum: mixin classes don't mix well with already mixed Enums (GH-9328) (GH-9486)

* bpo-29577: allow multiple mixin classes

(found with git bisect and your test)

[matrixise]

git bisect start
# bad: [1596fea] bpo-34899: Fix a possible assertion failure due to int_from_bytes_impl() (GH-9705)
git bisect bad 1596fea
# good: [84b0129] _Py_CoerceLegacyLocale() restores LC_CTYPE on fail (GH-9044) (GH-9046)
git bisect good 84b0129
# good: [470a435] bpo-34623: Use XML_SetHashSalt in _elementtree (GH-9146)
git bisect good 470a435
# bad: [92ad261] bpo-1529353: IDLE: Squeezer What's New for 3.6.7 (GH-9567)
git bisect bad 92ad261
# good: [e5fde1f] bpo-34537: Fix test_gdb:test_strings with LC_ALL=C (GH-9483)
git bisect good e5fde1f
# bad: [394e55a] [3.7] bpo-17239: Disable external entities in SAX parser (GH-9217) (GH-9511)
git bisect bad 394e55a
# bad: [44989bc] bpo-34472: Add data descriptor signature to zipfile (GH-8871) (GH-9399)
git bisect bad 44989bc
# good: [c00f703] bpo-34759: Fix error handling in ssl 'unwrap()' (GH-9468)
git bisect good c00f703
# bad: [0c076ca] [3.7] bpo-29577: Enum: mixin classes don't mix well with already mixed Enums (GH-9328) (GH-9486)
git bisect bad 0c076ca
# first bad commit: [0c076ca] [3.7] bpo-29577: Enum: mixin classes don't mix well with already mixed Enums (GH-9328) (GH-9486)

[ethanfurman]

New changeset cd45385 by Ethan Furman in branch 'master':
bpo-34909: keep searching mixins until base class is found (GH-9737)
cd45385

[ethanfurman]

New changeset 453b3b0 by Ethan Furman (Miss Islington (bot)) in branch '3.7':
bpo-34909: keep searching mixins until base class is found (GH-9737) (GH-9738)
453b3b0

[matrixise]

Ethan,

I think you have fixed the issue, could you close this issue?

Thank you

[ethanfurman]

Stéphane, thanks for the tip about bisect!

[ned-deily]

Can you please merge a NEWS item for this issue using blurb since it is a user-visible problem? I'll cherry-pick it into the 3.7.1 final.

[ethanfurman]

On 10/12/2018 09:52 PM, Ned Deily wrote:

> Can you please merge a NEWS item for this issue using blurb since it
> is a user-visible problem? I'll cherry-pick it into the 3.7.1 final.

I don't believe the problem exists in 3.7.0 -- but I also don't know how
to select the 3.7.0 "branch"/"bookmark"/"tag"/"whatever-it's-called" to
build and test myself.

If you can tell me how to get to whatever commit represents 3.7.0 I'm
happy to test.

[ned-deily]

My concern was that it seemed like at least one user had run into this while running 3.7.1rc1 (?), if I understood your comment; if so, I think we should mention in the changelog that the problem was fixed in rc2.

In any case, the state of the code for any release is captured as a git tag as described in the devguide. As always, when switching versions, it's important to start with a clean build directory (make distclean or git clean -fdxq may help).

git checkout v3.7.0

https://devguide.python.org/devcycle/#development-cycle

[ethanfurman]

Working on getting that news entry.

[ethanfurman]

Okay, the patch is here:

https://github.com/ethanfurman/cpython/tree/enum_news_entry

But I managed to screw something up, and my wife just got hit by a car so this is now a really low priority.

Ned, if you could grab the relevant pieces from the wreck that is that branch I would be grateful.

[ned-deily]

New changeset 8c9fd9c by Ned Deily in branch 'master':
bpo-34909: NEWS entry (GH-9995)
8c9fd9c

[matrixise]

Hi Ethan and Ned,

I am not fluent with the code of the Enum, but maybe I could work on this issue and try to finish it.

Ethan, I am really sorry for your wife, take care of her.

[ned-deily]

Thanks, Stéphane, but I’ve already taken care of it. I’ll be closing the issue shortly. (And ditto on the best wishes, Ethan.)

[matrixise]

Hi Ned, good news (I hope the best for the wife of Ethan)

[ned-deily]

New changeset 03ca8b5 by Ned Deily in branch '3.7':
bpo-34909: NEWS entry.
03ca8b5

[ned-deily]

NEWS entry added for 3.7.1 and master. Closing again.

[ethanfurman]

Thank you both.