[issue39113] PyUnicode_AsUTF8AndSize Sometimes Segfaults With Incomplete Surrogate Pair
New submission from william.ayd :
With the attached extension module, if I run the following in the REPL:
>>> import libtest
>>>
>>> libtest.error_if_not_utf8("foo")
'foo'
>>> libtest.error_if_not_utf8("\ud83d")
Traceback (most recent call last):
File "", line 1, in
UnicodeEncodeError: 'utf-8' codec can't encode character '\ud83d' in position
0: surrogates not allowed
>>> libtest.error_if_not_utf8("foo")
'foo'
Things seem OK. But the next invocation of
>>> libtest.error_if_not_utf8("\ud83d")
Then causes a segfault. Note that the order of the input seems important;
simply repeating the call with the invalid surrogate doesn't cause the segfault
--
files: testmodule.c
messages: 358755
nosy: william.ayd
priority: normal
severity: normal
status: open
title: PyUnicode_AsUTF8AndSize Sometimes Segfaults With Incomplete Surrogate
Pair
Added file: https://bugs.python.org/file48798/testmodule.c
___
Python tracker
<https://bugs.python.org/issue39113>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue39113] PyUnicode_AsUTF8AndSize Sometimes Segfaults With Incomplete Surrogate Pair
william.ayd added the comment: Hmm my mistake - thanks! -- ___ Python tracker <https://bugs.python.org/issue39113> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue35041] urllib.parse.quote safe Parameter Not Optional
New submission from william.ayd :
The safe parameter in urllib.parse.quote is documented as optional. However,
the following will raise TypeError: 'NoneType' object is not iterable:
urllib.parse.quote("/", safe=None)
whereas explicitly providing an iterable will allow the function to succeed:
urllib.parse.quote("/", safe=[])
--
messages: 328229
nosy: william.ayd
priority: normal
severity: normal
status: open
title: urllib.parse.quote safe Parameter Not Optional
versions: Python 3.7
___
Python tracker
<https://bugs.python.org/issue35041>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue35041] urllib.parse.quote safe Parameter Not Optional
william.ayd added the comment:
Semantics aside is it still the intended behavior that these calls should work:
urllib.parse.quote("/", safe='')
AND
urllib.parse.quote("/", safe=[])
But that this should raise?
urllib.parse.quote("/", safe=None)
IMO seems counterintuitive
--
___
Python tracker
<https://bugs.python.org/issue35041>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue35041] urllib.parse.quote safe Parameter Not Optional
william.ayd added the comment:
Hmm well I still personally feel that the implementation is somewhat off mores
than the documentation. Specifically I think it is confusing that it accepts an
empty iterable but not one containing elements.
This is fine:
urllib.parse.quote("/", safe=[])
Though this isn't:
urllib.parse.quote("/", safe=['/'])
Even though the following two calls are fine (though with different return
values as expected):
urllib.parse.quote("/", safe='')
urllib.parse.quote("/", safe='/')
It might go against the spirit of duck typing but I find it very nuanced that
empty iterables are allowed but if non-empty it must be a string. Would it not
make more sense to raise if a non-String type is passed?
--
___
Python tracker
<https://bugs.python.org/issue35041>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue35041] urllib.parse.quote safe Parameter Not Optional
william.ayd added the comment: What if we instead just raised for anything that isn't a string or a byte? The docstring for quote suggests that it should only accept str or byte objects for safe, though it doesn't enforce that: https://github.com/python/cpython/blob/121eb1694cab14df857ba6abe9839654cada15cf/Lib/urllib/parse.py#L791 Seems like it would be easier to enforce that rather than trying to accept any arbitrary iterable. -- ___ Python tracker <https://bugs.python.org/issue35041> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
