audioop extension should be enabled on 64-bit archs

[doko42]

BPO	993173
Nosy	@loewis, @doko42
Files	setup.diff

^{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 2004-07-19.16:52:08.000>
created_at = <Date 2004-07-18.07:16:07.000>
labels = ['build']
title = 'audioop extension should be enabled on 64-bit archs'
updated_at = <Date 2004-07-19.16:52:08.000>
user = 'https://github.com/doko42'

bugs.python.org fields:

activity = <Date 2004-07-19.16:52:08.000>
actor = 'loewis'
assignee = 'none'
closed = True
closed_date = None
closer = None
components = ['Build']
creation = <Date 2004-07-18.07:16:07.000>
creator = 'doko'
dependencies = []
files = ['6094']
hgrepos = []
issue_num = 993173
keywords = ['patch']
message_count = 2.0
messages = ['46387', '46388']
nosy_count = 2.0
nosy_names = ['loewis', 'doko']
pr_nums = []
priority = 'normal'
resolution = 'accepted'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue993173'
versions = ['Python 2.4']

[doko42]

[ forwarded from http://bugs.debian.org/260051 ]

Patch submitter writes:

The audioop extension is disabled on 64-bit
architectures, but the reasoning used in the
justification is flawed; it checks for whether "maxint"
is a 64-bit value, which it is, but audioop explicitly
uses a 32-bit integer type everywhere.

The attached patch re-enables this extension; tested on
alpha-linux, works fine.

The imageop and rgbimg extensions may also be usable on
64-bit platforms, but I haven't tested them.

[loewis]

Logged In: YES
user_id=21627

Thanks for the patch, applied as

setup.py 1.194
Setup.dist 1.46

Notice that the test for MAXINT is not a justification, but
the test itself to check for 64-bitness. The claim that the
module does not work on 64-bit systems originates from 1.65
of Modules/Setup.in. Unfortunately, Guido's claim that it
doesn't work back then is about as convincing as the Steve
Langasek's claim that it does work now.

If we find it does break, we should document more clearly
(or, perhaps, even fix it...)