PyEval_AcquireLock() and PyEval_AcquireThread() do not handle runtime finalization properly.

[ericsnowcurrently]

BPO	36475
Nosy	@vstinner, @ericsnowcurrently, @nanjekyejoannah
PRs	bpo-36475: Finalize PyEval_AcquireLock() and PyEval_AcquireThread() properly #12667 bpo-36475: Make PyThread_exit_thread with _Py_NO_RETURN #13068

^{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 2019-04-29.09:23:39.689>
created_at = <Date 2019-03-29.19:26:20.924>
labels = ['interpreter-core', 'type-bug', '3.8']
title = 'PyEval_AcquireLock() and PyEval_AcquireThread() do not handle runtime finalization properly.'
updated_at = <Date 2020-03-09.10:53:27.812>
user = 'https://github.com/ericsnowcurrently'

bugs.python.org fields:

activity = <Date 2020-03-09.10:53:27.812>
actor = 'vstinner'
assignee = 'none'
closed = True
closed_date = <Date 2019-04-29.09:23:39.689>
closer = 'vstinner'
components = ['Interpreter Core']
creation = <Date 2019-03-29.19:26:20.924>
creator = 'eric.snow'
dependencies = []
files = []
hgrepos = []
issue_num = 36475
keywords = ['patch']
message_count = 7.0
messages = ['339138', '339365', '339876', '341054', '341055', '341389', '363714']
nosy_count = 3.0
nosy_names = ['vstinner', 'eric.snow', 'nanjekyejoannah']
pr_nums = ['12667', '13068']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue36475'
versions = ['Python 3.8']

[ericsnowcurrently]

Daemon threads keep running until they finish or until finalization starts. For the latter, there is a check right after the thread acquires the GIL which causes the thread to exit if runtime finalization has started. [1] However, there are functions in the C-API that facilitate acquiring the GIL, but do not cause the thread to exit during finalization:

 PyEval_AcquireLock()
 PyEval_AcquireThread()

Daemon threads that acquire the GIL through these can cause a deadlock during finalization. (See issue bpo-36469.) They should probably be updated to match what PyEval_RestoreThread() does.

[1] see PyEval_RestoreThread() and the eval loop, in PyEval_EvalFrameEx()

[nanjekyejoannah]

@eric.snow , you can review the PR I submitted for this.

[nanjekyejoannah]

@eric do we need any tests for this?

[vstinner]

New changeset f781d20 by Victor Stinner (Joannah Nanjekye) in branch 'master':
bpo-36475: Finalize PyEval_AcquireLock() and PyEval_AcquireThread() properly (GH-12667)
f781d20

[vstinner]

I am not comfortable to backport this change to Python 3.7. It's too early to know how it will impact applications and how many complains we will get :-) If someone really wants to backport this scary change to 3.7, I would suggest to wait for 1 month after Python 3.8.0 final release.

I close the issue.

See bpo-36479 for the follow-up.

[vstinner]

New changeset c664b34 by Victor Stinner in branch 'master':
bpo-36475: Make PyThread_exit_thread with _Py_NO_RETURN (GH-13068)
c664b34

[vstinner]

I marked bpo-23592 as duplicate of this issue.