Remove distributed copy of libffi

[benjaminp]

BPO	12081
Nosy	@loewis, @birkenfeld, @doko42, @jcea, @amauryfa, @abalkin, @pitrou, @benjaminp, @merwok, @briancurtin, @meadori
Superseder	bpo-15194: libffi-3.0.11 update

^{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 2012-07-08.17:39:48.546>
created_at = <Date 2011-05-15.15:37:05.563>
labels = ['ctypes']
title = 'Remove distributed copy of libffi'
updated_at = <Date 2012-07-08.17:39:48.545>
user = 'https://github.com/benjaminp'

bugs.python.org fields:

activity = <Date 2012-07-08.17:39:48.545>
actor = 'loewis'
assignee = 'none'
closed = True
closed_date = <Date 2012-07-08.17:39:48.546>
closer = 'loewis'
components = ['ctypes']
creation = <Date 2011-05-15.15:37:05.563>
creator = 'benjamin.peterson'
dependencies = []
files = []
hgrepos = []
issue_num = 12081
keywords = []
message_count = 11.0
messages = ['136034', '136035', '136049', '136050', '136659', '158313', '158325', '158430', '158544', '165027', '165029']
nosy_count = 15.0
nosy_names = ['loewis', 'georg.brandl', 'doko', 'jcea', 'amaury.forgeotdarc', 'belopolsky', 'pitrou', 'nadeem.vawda', 'benjamin.peterson', 'eric.araujo', 'rpetrov', 'Arfrever', 'brian.curtin', 'meador.inge', 'rosslagerwall']
pr_nums = []
priority = 'normal'
resolution = 'duplicate'
stage = None
status = 'closed'
superseder = '15194'
type = None
url = 'https://bugs.python.org/issue12081'
versions = ['Python 3.3']

[benjaminp]

I believe the bugs which the patched version of libffi used have been fixed in recent versions. We should stop distributing an old version.

[birkenfeld]

Sounds reasonable. How will this work on Windows?

[rpetrov]

On windows work with patched version of library . Unpatched does not work but I cannot remember python issue number.

[Arfrever]

Has the patch been sent to libffi upstream? What was the response from libffi upstream?

[doko42]

iirc after merging 3.0.9, we still had to use the internal libffi bits for windows and macosx. I didn't check 3.0.10rc8

[rosslagerwall]

In any case, it should be OK to remove libffi_arm_wince?

Is WinCE supported?

[pitrou]

I don't think we have ever "supported" WinCE (which is apparently named "Windows Embedded Compact 7" nowadays). It only provides a subset of the Win32 API so the current tree may not even compile.

[loewis]

Arfrever: I doubt anybody has contributed patches back, or that anybody is interested in doing so.

I personally don't see a problem in using an old libffi version, so I fail to see Benjamin's issue. Figuring out how exactly to use the system libffi is more hassle than keeping our own copy.

Please understand that ctypes is unmaintained. Anybody actively taking over maintenance of ctypes would have to decide on how integration with libffi is supposed to work. Without a maintainer, falling back to the sytem libffi is a too high risk, IMO, since this will certainly produce tons of new bug reports, with nobody prepared to deal with them.

[doko42]

The last time I merged libffi, we were not able to build the MacOS X and Windows libffi from the upstream sources, but used the internal copy of the copy. Now that libffi 3.0.11 is released, we could

• update to this new version, see if the MacOS X and Windows
  builds work (there are upstream related changes)

• start defaulting to the system libffi for at least some targets
  (e.g. linux), for which we know that almost nobody will use the
  internal copy

No, I don't volunteer to maintain ctypes itself.

[meadori]

Matthias recently updated libffi to 3.0.11 (bpo-15194). It would seem that we intend to keep a local copy of the libffi sources for now and that this issue can be closed. Does anyone see a reason to keep this open?

[loewis]

Closing as a duplicate. The original issue is resolved: we are not distributing an old copy of libffi anymore.