Repository navigation
test_json segfault on OpenBSD #69529
Description
| BPO | 25342 |
|---|---|
| Nosy | @vstinner, @skrah, @serhiy-storchaka, @matrixise |
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 = None
created_at = <Date 2015-10-08.11:57:39.664>
labels = []
title = 'test_json segfault on OpenBSD'
updated_at = <Date 2015-10-10.15:51:58.146>
user = 'https://bugs.python.org/rpointel'bugs.python.org fields:
activity = <Date 2015-10-10.15:51:58.146>
actor = 'skrah'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = []
creation = <Date 2015-10-08.11:57:39.664>
creator = 'rpointel'
dependencies = []
files = []
hgrepos = []
issue_num = 25342
keywords = []
message_count = 12.0
messages = ['252531', '252532', '252533', '252534', '252535', '252537', '252538', '252539', '252540', '252547', '252550', '252721']
nosy_count = 5.0
nosy_names = ['vstinner', 'skrah', 'rpointel', 'serhiy.storchaka', 'matrixise']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = None
url = 'https://bugs.python.org/issue25342'
versions = ['Python 3.6']
Activity
matrixise commentedon Oct 8, 2015
What the details of this issue?
On 8 Oct 2015, at 13:57, Remi Pointel wrote:
rpointel commentedon Oct 8, 2015
Sorry I clicked on "create" before adding info in comment.
When I run the test_json suite on OpenBSD, I have a segfault:
[-]json[/-][+]test_json segfault on OpenBSD[/+]rpointel commentedon Oct 8, 2015
It fails on test_endless_recursion:
test_endless_recursion (test.test_json.test_recursion.TestPyRecursion) ... zsh: segmentation fault (core dumped)
skrah commentedon Oct 8, 2015
Try setting the stack size to 8MB (the default on Linux) in login.conf.
rpointel commentedon Oct 8, 2015
It's good when I entered "ulimit -s 8192". Sorry for the noise...
skrah commentedon Oct 8, 2015
It's not noise, we also have problems on Windows with test_json.
In theory we should set Py_DEFAULT_RECURSION_LIMIT in Python/ceval.c
to appropriate values for the default OS stack sizes in order to get
a proper RuntimeError instead of a segfault.
For OpenBSD it's clearly too high for the default. Could you try
to find a safe value (perhaps 500 instead of 1000) for OpenBSD's
default?
vstinner commentedon Oct 8, 2015
Would it be possible to compute the recursion limit depending on the current maximum C stack size? At least estimate it? I guess the sys.getdefaultlimit() is 1000 is an arbitrary value. I also know that the effective limit depends on the memory allocated on the stack by C functions. Calling a Python function and a function implemented in C is different :-/
skrah commentedon Oct 8, 2015
The ratio (1000/8MB) seems to be stable on Linux. Perhaps we could
write a getrlimit program for ./configure to get the stack size and
adjust the recursion limit to keep the ratio the same.
It's just an estimate of course.
skrah commentedon Oct 8, 2015
Or did you want to compute it at runtime? Actually, why not?
vstinner commentedon Oct 8, 2015
Stefan Krah added the comment:
I'm asking to adjust the limit at _runtime_. getrlimit() is not
something static, it's common to modify them manually for the current
shell process (and child process), or system-wide.
serhiy-storchaka commentedon Oct 8, 2015
Duplicates and related issues: bpo-25329, bpo-25222, bpo-24999, bpo-22984, bpo-18075, bpo-12980.
skrah commentedon Oct 10, 2015
Adjusting the estimate at runtime sounds good to me.
kumaraditya303 commentedon Jun 19, 2022
Closing as OpenBSD is not supported as per PEP 11.
vstinner commentedon Jun 20, 2022
Lib/test/test_json/test_recursion.py now uses
with support.infinite_recursion():which uses a lower recursion limit and so is less likely to crash with a stack overflow. Previously, the test required a large stack.