bpo-47000: Make io.text_encoding() respects UTF-8 mode
#32003
Conversation
| @@ -479,7 +480,13 @@ _io_text_encoding_impl(PyObject *module, PyObject *encoding, int stacklevel) | |||
| return NULL; | |||
| } | |||
| } | |||
| return &_Py_ID(locale); | |||
There was a problem hiding this comment.
@ericsnowcurrently We need to incref this until we implement immortal object. Am I correct?
There was a problem hiding this comment.
Sort of. Currently all statically allocated global objects have their refcount initialized to a really large value. So we haven't been bothering with the incref if we know the object is one of those, which it is in this case. Incref'ing it won't hurt. However, it is effectively unnecessary.
There was a problem hiding this comment.
I am afraid breaking refleak buildbot.
Would you tell my why this return &_Py_ID(locale); don't break refleak test?
There was a problem hiding this comment.
The refcount of static objects is not directly included in _Py_RefTotal, which is used for refleak detection. (_Py_RefTotal will still ensure that refcount operations are balanced though, regardless of the objects involved.)
There was a problem hiding this comment.
(_Py_RefTotal will still ensure that refcount operations are balanced though, regardless of the objects involved.)
So skipping INCREF here might cause in-balance of _Py_RefTotal. Doesn't it break refleak test?
There was a problem hiding this comment.
Good point. It didn't cause a problem before. I'd have to look into why not.
|
When you're done making the requested changes, leave the comment: |
Co-authored-by: Eric Snow <[email protected]>
In a gh-32003 comment, I realized it wasn't very clear how _Py_DECLARE_STR() should be used. This changes adds a comment to clarify. https://bugs.python.org/issue46541
Co-authored-by: Eric Snow <[email protected]>
methane
changed the title
io.text_encoding() respects UTF-8 mode
https://bugs.python.org/issue47000