timezone allows no offset from range (23:59, 24:00)

[joernheissler]

BPO	37642
Nosy	@abalkin, @pganssle, @joernheissler, @tirkarthi
PRs	bpo-37642: Update max and min offset in datetime module #14878 [3.7] bpo-37642: Update max and min offset in datetime module (GH-14878) #15226 [3.8] bpo-37642: Update max and min offset in datetime module (GH-14878) #15227

^{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 = None
closed_at = <Date 2019-08-15.19:11:42.809>
created_at = <Date 2019-07-21.11:41:24.145>
labels = ['3.7', '3.8', 'type-bug', 'library', '3.9']
title = 'timezone allows no offset from range (23:59, 24:00)'
updated_at = <Date 2019-08-15.19:11:42.808>
user = 'https://github.com/joernheissler'

bugs.python.org fields:

activity = <Date 2019-08-15.19:11:42.808>
actor = 'p-ganssle'
assignee = 'none'
closed = True
closed_date = <Date 2019-08-15.19:11:42.809>
closer = 'p-ganssle'
components = ['Library (Lib)']
creation = <Date 2019-07-21.11:41:24.145>
creator = 'joernheissler'
dependencies = []
files = []
hgrepos = []
issue_num = 37642
keywords = ['patch']
message_count = 6.0
messages = ['348237', '348242', '348298', '349290', '349816', '349817']
nosy_count = 4.0
nosy_names = ['belopolsky', 'p-ganssle', 'joernheissler', 'xtreak']
pr_nums = ['14878', '15226', '15227']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue37642'
versions = ['Python 3.7', 'Python 3.8', 'Python 3.9']

[joernheissler]

https://bugs.python.org/issue5288 changed datetime.timezone to accept sub-minute offsets.

The C implementation allows offsets from range (23:59, 24:00) while the python implementation does not:

# C
>>> timezone(timedelta(seconds=86399))
datetime.timezone(datetime.timedelta(seconds=86399))

# Python
>>> timezone(timedelta(seconds=86399))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "cpython/Lib/datetime.py", line 2194, in __new__
    raise ValueError("offset must be a timedelta "
ValueError: offset must be a timedelta strictly between -timedelta(hours=24) and timedelta(hours=24).

This is because _maxoffset is defined as timedelta(hours=23, minutes=59)

Second issue: The (undocumented) "min" and "max" attributes (both C and python) show 23:59
even though the C implementation can get closer to 24:00.
Should this be changed to timezone(timedelta(seconds=86399, microseconds=999999))?

(Same applies to the minimums)

[pganssle]

I agree that the C and Python behavior should be the same, and both of them should allow all offsets less than 24 hours. I'm actually quite surprised we don't have a test for this in datetimetester.py.

[pganssle]

I have been thinking about this more and I think the two issues (min and max vs. the incompatibility between C and Python) should be considered two separate issues.

My reasoning is that a change of the value of timezone.min and timezone.max is not something I'd be comfortable backporting to 3.7, because it has the potential to cause failures in some test suites (for example, datetime.now(tz=datetime.timezone.max).isoformat() will start returning a string that does not conform to ISO 8601, which has no support for sub-minute offsets). However, differences between the C and Python implementations are a violation of at least the spirit of PEP-399 and I think it should be backported to 3.7.

[pganssle]

New changeset 92c7e30 by Paul Ganssle (Ngalim Siregar) in branch 'master':
bpo-37642: Update acceptable offsets in timezone (GH-14878)
92c7e30

[pganssle]

New changeset 27b38b9 by Paul Ganssle in branch '3.8':
bpo-37642: Update acceptable offsets in timezone (GH-14878) (bpo-15227)
27b38b9

[pganssle]

New changeset ed44b84 by Paul Ganssle in branch '3.7':
bpo-37642: Update acceptable offsets in timezone (GH-14878) (bpo-15226)
ed44b84