asyncore sillies

[mwhudson]

BPO	971200
Nosy	@mwhudson, @tim-one

^{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 2004-06-14.10:36:08.000>
created_at = <Date 2004-06-11.15:19:06.000>
labels = []
title = 'asyncore sillies'
updated_at = <Date 2004-06-14.10:36:08.000>
user = 'https://github.com/mwhudson'

bugs.python.org fields:

activity = <Date 2004-06-14.10:36:08.000>
actor = 'mwh'
assignee = 'none'
closed = True
closed_date = None
closer = None
components = ['None']
creation = <Date 2004-06-11.15:19:06.000>
creator = 'mwh'
dependencies = []
files = []
hgrepos = []
issue_num = 971200
keywords = []
message_count = 7.0
messages = ['21142', '21143', '21144', '21145', '21146', '21147', '21148']
nosy_count = 3.0
nosy_names = ['mwh', 'tim.peters', 'anthonybaxter']
pr_nums = []
priority = 'high'
resolution = None
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue971200'
versions = []

[mwhudson]

current cvs head, mac os x 10.2, debug build of python.

test_asynchat fails if and only if the compiled asyncore.pyc
file is present.

this makes no sense to me, but it's consistent.

[mwhudson]

Logged In: YES
user_id=6656

Oh my:

>>> 1e309
Inf
[40577 refs]
>>> marshal.loads(marshal.dumps(1e309))
0.0
[40577 refs]

this must be the new "LC_NUMERIC agnostic" stuff, right?

[tim-one]

Logged In: YES
user_id=31435

Well, what marshal (or pickle) do with an infinity (or NaN, or
the sign of a signed zero) is a platform accident. Here with
the released 2.3.4 on Windows (which doesn't have any
LC_NUMERIC changes):

Python 2.3.4 (#53, May 25 2004, 21:17:02) [MSC v.1200 32 
bit (Intel)] on win32
Type "help", "copyright", "credits" or "license" for more 
information.
>>> 1e309
1.#INF
>>> import marshal
>>> marshal.loads(marshal.dumps(1e309))
1.0
>>>

So simply can't use a literal 1e309 in compiled code. There's
no portable way to spell infinity in Python. PEP-754 would
introduce a reasonably portable way, were it accepted.
Before then, 1e200*1e200 is probably the easiest reasonably
portable way -- but since behavior in the presence of an
infinity is accidental anyway, much better to avoid using
infinity at all in the libraries.

[mwhudson]

Logged In: YES
user_id=6656

actually, i think the summary is that the most recent change to
asyncore is just broken. blaming the recent changes around
LC_NUMERIC and their effect or non-effect on marshal was a read
herring.

[anthonybaxter]

Logged In: YES
user_id=29957

Is it worth making marshal issue a warning when someone
tries to store an infinity, in that case? It could (ha!) be
suprising that a piece of code that works from a .py fails
silently from a .pyc.

[tim-one]

Logged In: YES
user_id=31435

marshal doesn't know whether the input is an infinity; that's
why the result is a platform-dependent accident; "infinity"
isn't a C89 concept, and Python inherits its ignorance from C

[mwhudson]

Logged In: YES
user_id=6656

I think this can be closed now. A half-assed test could be
added to marshal (if (x!=x) PyErr_SetString(...)) but I'm
not sure it's worth it.