asyncio: Process must close the transport when the process exit

[vstinner]

BPO	23242
Nosy	@gvanrossum, @vstinner, @agronholm, @1st1
Files	subprocess_close.patch read_subprocess.py

^{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 2015-01-15.13:36:57.010>
created_at = <Date 2015-01-14.23:11:18.655>
labels = ['expert-asyncio']
title = 'asyncio: Process must close the transport when the process exit'
updated_at = <Date 2017-01-07.19:58:07.672>
user = 'https://github.com/vstinner'

bugs.python.org fields:

activity = <Date 2017-01-07.19:58:07.672>
actor = 'alex.gronholm'
assignee = 'none'
closed = True
closed_date = <Date 2015-01-15.13:36:57.010>
closer = 'vstinner'
components = ['asyncio']
creation = <Date 2015-01-14.23:11:18.655>
creator = 'vstinner'
dependencies = []
files = ['37707', '46198']
hgrepos = []
issue_num = 23242
keywords = ['patch']
message_count = 4.0
messages = ['234042', '234070', '234075', '284930']
nosy_count = 5.0
nosy_names = ['gvanrossum', 'vstinner', 'alex.gronholm', 'python-dev', 'yselivanov']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue23242'
versions = ['Python 3.4', 'Python 3.5', 'Python 3.6']

[vstinner]

The asyncio.subprocess.Process class never closes the subprocess transport, whereas this transport is private.

I propose to close the transport as soon as possible: when transport.get_returncode() is called and its result is not None.

I'm not sure that it's the most efficient way to close the transport. It may be better to close the transport in the connection_lost() method of the protocol.

The patch also checks in Process.communicate() that the process is alive, and is clears the reference to the transport.

[python-dev]

New changeset df493e9c6821 by Victor Stinner in branch '3.4':
Issue bpo-23242: asyncio.SubprocessStreamProtocol now closes the subprocess
https://hg.python.org/cpython/rev/df493e9c6821

[vstinner]

I'm not sure that it's the most efficient way to close the transport. It may be better to close the transport in the connection_lost() method of the protocol.

"process_exited", not "connection_lost".

I implemented this option which is simpler and more efficient (always and immediatly close the transport).

Clearing the reference to the transport in the Process is less important, it can be done later in a different issue/changeset. I close this issue.

[agronholm]

Are you sure this has been fixed? The attached script reproduces the bug 100% reliably on my laptop.