The following code raise a ValueError('I/O operation on closed file.'):
---
import socket
socket.socket(socket.SOCK_STREAM, socket.AF_INET)
s=socket.socket(socket.SOCK_STREAM, socket.AF_INET)
f=s.makefile("rb")
f.close()
print(repr(f))
---

io.BufferedReader.__repr__() reads self.name (self.buffer.name), but self.buffer is closed.

io.BufferedReader.__repr__() catchs AttributeError when reading self.name. It should also maybe catch ValueError: attached patch does that.

socket.repr(x) returns a string with "[closed]". BufferedReader.repr() should maybe do the same.

Note: TextIOWrapper has the same issue.

Antoine suggested me to patch SocketIO.name property instead of the BufferedReader.__repr__() method: socketio_name.patch implements this idea.

Fixed by r87730.

I find this new behavior a usability regression. Before this change, code
(e.g python 2 code ported to python 3) could do:

   fd = sock.fileno()

Without handling errors, since closed socket would raise (good). Now such code need to check the return value (bad):

   fd = sock.fileno()
   if fd == -1:
       fail...

This is also not consistent with other objects:

>>> f = open("Makefile")
>>> f.fileno()
3
>>> f.close()
>>> f.fileno()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: I/O operation on closed file
>>> repr(f)
"<_io.TextIOWrapper name='Makefile' mode='r' encoding='UTF-8'>"


The issue with repr() on closed socket can be mitigated easily inside __repr__, handling closed sockets without affecting code using file descriptors.

Can we return the old safe behavior?

This issue is closed. Please open a new issue.