Race condition in pexpect.spawn() on Solaris 11

[reynir]

It seems it is possible in certain situations for the child_fd to be closed before the windowsize can be set. I tried putting in a print before, and the file descriptor looks OK (i.e. it's not -1).

import pexpect
pexpect.spawn("/bin/true")

Relevant stack trace:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/export/home/rbj/.local/lib/python2.6/site-packages/pexpect/__init__.py", line 511, in __init__
    self._spawn(command, args)
  File "/export/home/rbj/.local/lib/python2.6/site-packages/pexpect/__init__.py", line 674, in _spawn
    self.setwinsize(24, 80)
  File "/export/home/rbj/.local/lib/python2.6/site-packages/pexpect/__init__.py", line 1602, in setwinsize
    fcntl.ioctl(self.fileno(), TIOCSWINSZ, s)
IOError: [Errno 6] No such device or address

This is a corner case that probably doesn't affect many users (if any). You can download a Solaris 11.2 VirtualBox image if you want to test it yourself. I have the CPU throttled to 50% on my VM - that might make the race condition more obvious.

[takluyver]

I guess if the child process finishes very quickly, the pty might be destroyed before the parent tries to act on it. We should probably catch and silence that errno.

[jquast]

Sadly this isn't reproducing on joyent SmartOS (OpenSolaris) with a 0.125 CPU share. I'll try the commercial Solaris, then.

[jquast]

this will be resolved in next pexpect release

[jquast]

Closed 2015-05-20 by release of ptyprocess 0.5, please open new issue if reproducing.