Repository navigation
segfaults calling warnings.warn() with non-string message #47889
Description
| BPO | 3639 |
|---|---|
| Nosy | @brettcannon, @devdanzin, @benjaminp |
| Files |
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/brettcannon'
closed_at = <Date 2008-09-02.04:02:12.729>
created_at = <Date 2008-08-21.22:30:26.257>
labels = ['interpreter-core', 'type-crash', 'release-blocker']
title = 'segfaults calling warnings.warn() with non-string message'
updated_at = <Date 2008-09-02.04:02:12.727>
user = 'https://bugs.python.org/dalcinl'bugs.python.org fields:
activity = <Date 2008-09-02.04:02:12.727>
actor = 'brett.cannon'
assignee = 'brett.cannon'
closed = True
closed_date = <Date 2008-09-02.04:02:12.729>
closer = 'brett.cannon'
components = ['Interpreter Core']
creation = <Date 2008-08-21.22:30:26.257>
creator = 'dalcinl'
dependencies = []
files = ['11211', '11220']
hgrepos = []
issue_num = 3639
keywords = ['patch']
message_count = 10.0
messages = ['71694', '71709', '71721', '71750', '71763', '71765', '71766', '71773', '71774', '72312']
nosy_count = 4.0
nosy_names = ['brett.cannon', 'dalcinl', 'ajaksu2', 'benjamin.peterson']
pr_nums = []
priority = 'release blocker'
resolution = 'accepted'
stage = None
status = 'closed'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue3639'
versions = ['Python 3.0']
Activity
dalcinl commentedon Aug 21, 2008
devdanzin commentedon Aug 22, 2008
Two small clues.
First, a backtrace:
That is, _warnings.warn(spam) works OK and avoids the
warnings.warn(spam) crash for values already called by the former.
brettcannon commentedon Aug 22, 2008
If you search for _PyUnicode_AsString() in Python/_warnings.c you will
find several places that assume that the proper measures have been taken
to make sure the object is a string. All of those places need to be
fixed so that if a string is not passed in then one is grabbed.
And the reason this turned out as a segfault is for a missing error
return value just before the strcmp() call.
devdanzin commentedon Aug 22, 2008
Brett,
I don't think I know C (and CPython) enough to fix this. I was able to
get rid of this specific segfault with this:
+ const char *text_char =
_PyUnicode_AsString(PyObject_Str(text));
But I have no idea whether I should also incref/decref the PyObject_Str.
brettcannon commentedon Aug 22, 2008
On Fri, Aug 22, 2008 at 8:03 AM, Daniel Diniz <report@bugs.python.org> wrote:
That's along the lines of what needs to be done (and what I was
planning on doing), although you need to do more error checking on the
return values. Plus the patch I am cooking up adds more checks in the
code for the return value of _PyUnicode_AsString().
brettcannon commentedon Aug 22, 2008
The patch doesn't actually bother with a translation as the code causing
issue is only there to prevent infinite recursion. So if the object
being used is not a string, then there is no need to worry as it is not
part of the infinite recursion problem.
I also added a bunch of missing error checks.
benjaminp commentedon Aug 22, 2008
Brett, is this patch ready for review?
brettcannon commentedon Aug 22, 2008
That's why the keyword is set. =)
benjaminp commentedon Aug 22, 2008
On Fri, Aug 22, 2008 at 2:59 PM, Brett Cannon <report@bugs.python.org> wrote:
Ah. I missed that. :) The patch looks fine.
--
Cheers,
Benjamin Peterson
"There's no place like 127.0.0.1."
brettcannon commentedon Sep 2, 2008
Checked in r66140.