Patch for bug 580952: import lock should be exposed

[jdhildeb]

BPO	683257
Nosy	@gvanrossum
Files	python2.2.2-import.patch: patch against 2.2.2 python2.3-2.2.99-import.patch: patch against 2.3-2.2.99

^{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 = 'https://github.com/gvanrossum'
closed_at = <Date 2003-02-13.17:11:47.000>
created_at = <Date 2003-02-09.04:22:26.000>
labels = ['interpreter-core']
title = 'Patch for bug 580952: import lock should be exposed'
updated_at = <Date 2003-02-13.17:11:47.000>
user = 'https://bugs.python.org/jdhildeb'

bugs.python.org fields:

activity = <Date 2003-02-13.17:11:47.000>
actor = 'gvanrossum'
assignee = 'gvanrossum'
closed = True
closed_date = None
closer = None
components = ['Interpreter Core']
creation = <Date 2003-02-09.04:22:26.000>
creator = 'jdhildeb'
dependencies = []
files = ['5009', '5010']
hgrepos = []
issue_num = 683257
keywords = ['patch']
message_count = 4.0
messages = ['42734', '42735', '42736', '42737']
nosy_count = 2.0
nosy_names = ['gvanrossum', 'jdhildeb']
pr_nums = []
priority = 'low'
resolution = 'accepted'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue683257'
versions = ['Python 2.3']

[jdhildeb]

See bug 580952 for all the details.

I've attached patches against Python 2.3-2.2.99 (it
would be great to have this in for 2.3 final), and also
against Python 2.2.2 (for the next 2.2.x maintenance
release).

[gvanrossum]

Logged In: YES
user_id=6380

I've applied the 2.3 patch, with some changes to avoid the
fatal error in unlock_import().

Open issues:

• This needs docs. Jason, can you provide a doc patch?

• Why do you want this backported to 2.2? It's a new feature
  after all. Jason, can you comment on that?

[jdhildeb]

Logged In: YES
user_id=173690

Yes, I can provide a doc patch. Is there anything that
needs to be updated other than Doc/lib/libimp.tex? (I'm new
to the python source tree)

Re: backporting: Here's my reasoning:

For a multithreaded Python app which needs to use an import
hook, there is no way to do this reliably without access to
the internal import lock. If the app uses a separate lock
it is possible for a deadlock to happen (see bug 580952 for
explanation), yet if no locking is done the runtime
environment can get really screwed up (I've seen it happen).

So from the perspective of such an app, the Python import
API is buggy/incomplete because it provides no safe way to
use an import hook. Therefore I view this patch more as a
fix than a new feature.

[gvanrossum]

Logged In: YES
user_id=6380

Neal Norwitz already add the docs and a test. :-)

Some belated feedback on your patch: when returning Py_None,
you must Py_INCREF(Py_None)! Also, please use trailing \ in
multi-line string literals; the C std requires this even
though GCC doesn't. :-) (Thanks Neal!)