Re: [Plash] Odd popen behaviour

[Thomas Leonard]

On 6/20/06, Toby Murray <address@hidden> wrote:
> I was able to track this down to the os.popen2() implementation (at
> least in pyton 2.4 on my Fedora Core 5 box here)
[...]
> You'll see in there that the child that's forked as part of the popen2()
> first closes all of its file descriptors from 3 through to 1023
> inclusive before doing anything much else. In so doing, it closes the
> file descriptor that's used to communicate with the plash "file server".
> This prevents it from executing properly from then on.
[...]
> That said, the above could all be bogus, since the plash doco at
> http://plash.beasts.org/environment.html specifically indicates that
> close() has been modifiefd to prevent one from losing the fd used to
> communicate with the plash server. If this is the case, perhaps this
> code is failing because the child may not have the PLASH_COMM_FD
> environment variable set. (Sorry, I'm not infront of my machine anymore
> so this is just speculation).

Looks like you're right. Here's a simpler version with the same problem:

import os
fd = os.environ.get('PLASH_COMM_FD', None)
print "FD:", fd
file('/etc/passwd')
if fd is not None:
        os.close(int(fd))
        print "Closed"
        file('/etc/passwd')

This produces:

$ python t.py
FD: None

$ pola-run --prog /usr/bin/python -B -a t.py -fw /
FD: 15
Closed
sendmsg: Bad file descriptor
recvmsg: Bad file descriptor
[14326] cap-protocol: [fd 15] to-server: connection error, errno 9
(Bad file descriptor)
Traceback (most recent call last):
 File "t.py", line 8, in ?
IOError: [Errno 38] Function not implemented: '/etc/passwd'

As far as I can see, plash's new_close() function isn't being called.


--
Dr Thomas Leonard               http://rox.sourceforge.net
GPG: 9242 9807 C985 3C07 44A6  8B9A AE07 8280 59A5 3CC1