HTTPError interface changes / breaks depending on what was passed to constructor

[Keto]

BPO	13567
Nosy	@ezio-melotti, @merwok, @berkerpeksag, @demianbrecht
Files	httperror-geturl-fix-2.patch: HTTPError patch for 2.7 httperror-geturl-fix-3.patch: HTTPError patch for 3.3

^{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 = None
created_at = <Date 2011-12-09.15:54:42.132>
labels = ['type-bug', 'library']
title = 'HTTPError interface changes / breaks depending on what was passed to constructor'
updated_at = <Date 2015-04-26.16:41:44.398>
user = 'https://bugs.python.org/Keto'

bugs.python.org fields:

activity = <Date 2015-04-26.16:41:44.398>
actor = 'demian.brecht'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2011-12-09.15:54:42.132>
creator = 'Keto'
dependencies = []
files = ['23892', '23893']
hgrepos = []
issue_num = 13567
keywords = ['patch']
message_count = 2.0
messages = ['149106', '149107']
nosy_count = 5.0
nosy_names = ['ezio.melotti', 'eric.araujo', 'berker.peksag', 'Keto', 'demian.brecht']
pr_nums = []
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue13567'
versions = ['Python 2.7', 'Python 3.4', 'Python 3.5']

[Keto]

In case of authentication error, HTTPError gets initialized without file object and constructor of addinfourl is not called. This means that url attribute is not set and geturl() (inherited from addinfourl) raises AttributeError.

geturl() is not documented as part of HTTPError interface, so I'm not sure if it is correct to expect it to work. And of course this can be worked around by using the HTTPError.filename which always contains the url passed to constructor.

How ever, I have made a simple patch to fix the missing attribute.

There is also bpo-5286 with a patch, which makes the http_error_auth_reqed pass the fp to HTTPError making it a file-like object as stated in the documentation. So that would probably be the correct solution.

IMHO it's a bit bad design, when objects interface changes depending on how it was initialized.

[Keto]

Patch adapted to 3.3

[devdanzin]

The patch seems to still apply. Given the 14 years without an update, setting to pending and will close unless we get feedback about the desirability of this change.

[merwok]

Bugs don’t become invalid just because they’re old. We need to review the severity of the bug and the impact of a fix (difficulty of fixing, easier or harder maintenance, worth having different behaviour between versions, new incompatibilities…).

(A general comment: Triage is useful! But just poking issues creates notifications for everyone (or maybe just me 😛) without substantially advancing the resolution. Maybe focus deeper on some issues?)