Re: Python 3 test failures

[Ludovic Courtès]

Mark H Weaver <address@hidden> skribis:

> address@hidden (Ludovic Courtès) writes:
>
>> Mark H Weaver <address@hidden> skribis:
>>
>>> ======================================================================
>>> ERROR: test_fork (test.test_pty.PtyTest)
>>> ----------------------------------------------------------------------
>>> Traceback (most recent call last):
>>>   File 
>>> "/tmp/nix-build-python-3.3.3.drv-13/Python-3.3.3/Lib/test/test_pty.py", 
>>> line 116, in test_fork
>>>     pid, master_fd = pty.fork()
>>>   File "/tmp/nix-build-python-3.3.3.drv-13/Python-3.3.3/Lib/pty.py", line 
>>> 107, in fork
>>>     master_fd, slave_fd = openpty()
>>>   File "/tmp/nix-build-python-3.3.3.drv-13/Python-3.3.3/Lib/pty.py", line 
>>> 29, in openpty
>>>     master_fd, slave_name = _open_terminal()
>>>   File "/tmp/nix-build-python-3.3.3.drv-13/Python-3.3.3/Lib/pty.py", line 
>>> 70, in _open_terminal
>>>     raise os.error('out of pty devices')
>>> OSError: out of pty devices
>>
>> This particular test is fixed by the daemon patch I posted, which makes
>> sure /dev/pts/ptmx is 0666 (I’ll update nix-upstream.)
>
> What are the security implications of this?  Why do you suppose that
> Debian (or Linux, the kernel?) uses perms 0000 for /dev/pts/ptmx, 

The devpts file system of Linux creates this file with perms 000 by
default:

  https://www.kernel.org/doc/Documentation/filesystems/devpts.txt

> and how it is that these tests normally work outside of guix-daemon,
> when the perms are 0000?

Often, outside the chroot, /dev/ptmx is directly a character device,
with perms 0666.  Conversely, inside the chroot it’s a symlink to
/dev/pts/ptmx (as suggested in devpts.txt), but until this patch was
applied, it had perms 000.

Ludo’.