Skip to content

bpo-31249: Fix ref cycle in ThreadPoolExecutor#3178

Merged
vstinner merged 4 commits into
python:masterfrom
vstinner:futures_cycle
Aug 22, 2017
Merged

bpo-31249: Fix ref cycle in ThreadPoolExecutor#3178
vstinner merged 4 commits into
python:masterfrom
vstinner:futures_cycle

Conversation

@vstinner

@vstinner vstinner commented Aug 22, 2017

Copy link
Copy Markdown
Member

WorkItem.run() of concurrent.futures used by ThreadPoolExecutor now
breaks a reference cycle between an exception object and the WorkItem
object.

ThreadPoolExecutor.shutdown(wait=True) now also clears the set of
threads.

https://bugs.python.org/issue31249

Comment thread Lib/concurrent/futures/thread.py Outdated

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This change is not strictly needed to fix the initial bug:
https://bugs.python.org/issue31249#msg300634

But it should help to prevent leaking dangling threads if the executor object lives longer than expected.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have the feeling you're getting obsessed about "dangling threads".

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@pitrou: I have the feeling you're getting obsessed about "dangling threads".

That's true. In my experience, sometimes, it shows a larger issue. For example, https://bugs.python.org/issue31249 pointed a reference cycle which wasn't seen before.

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm working on race conditions since 3-6 months, and I saw so many non deterministic behaviours. I would like to make Python test suite more reliable, more reproductible.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At least, this part should be under the if wait clause IMO.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also, there is no race condition here. Besides the Thread object doesn't seem to hold onto any important data, so it shouldn't be a problem if isn't collected immediately.

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At least, this part should be under the if wait clause IMO.

I hesitated to not clear threads on shutdown(wait=False), but I don't see how clearing _theads can introduce any issue.

Anyway, I made the change ;-)

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also, there is no race condition here. Besides the Thread object doesn't seem to hold onto any important data, so it shouldn't be a problem if isn't collected immediately.

Hum, you know what? I reverted my changes on shutdown() to only break the reference cycle in _WorkItem ;-) Let's keep the change as small as possible.

concurrent.futures: WorkItem.run() used by ThreadPoolExecutor now
breaks a reference cycle between an exception object and the WorkItem
object. ThreadPoolExecutor.shutdown() now also clears its threads
set.
@vstinner

Copy link
Copy Markdown
Member Author

@gpshead, @pitrou: What do you think of the added "self._threads.clear()" statement?

@vstinner vstinner merged commit bc61315 into python:master Aug 22, 2017
@vstinner vstinner deleted the futures_cycle branch August 22, 2017 14:50
@gpshead

gpshead commented Aug 29, 2017

Copy link
Copy Markdown
Member

given I didn't respond earlier and I see no more self._threads.clear() in the accepted change, this change looks good to me. 👍

@Mariatta

Mariatta commented Sep 6, 2017

Copy link
Copy Markdown
Member

Already backported in #3253 so I'm removing the label.

GadgetSteve pushed a commit to GadgetSteve/cpython that referenced this pull request Sep 10, 2017
* bpo-31249: Fix ref cycle in ThreadPoolExecutor

concurrent.futures: WorkItem.run() used by ThreadPoolExecutor now
breaks a reference cycle between an exception object and the WorkItem
object. ThreadPoolExecutor.shutdown() now also clears its threads
set.

* shutdown() now only clears threads if wait is true.

* Revert changes on shutdown()
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants