[13671] in Athena Bugs
Re: Emacs/pty handling
daemon@ATHENA.MIT.EDU (Sam Hartman)
Mon Jul 17 13:24:20 1995
To: "Richard Basch" <basch@MIT.EDU>
Cc: bugs@MIT.EDU, rs6kswat@MIT.EDU
In-Reply-To: Your message of "Mon, 17 Jul 1995 13:09:04 EDT."
<9507171709.AA15548@tardis.MIT.EDU>
Date: Mon, 17 Jul 1995 13:24:16 EDT
From: Sam Hartman <hartmans@MIT.EDU>
>>>>> "Richard" == Richard Basch <basch@MIT.EDU> writes:
Richard> Anyway, if you start emacs from a remote login, and then
Richard> the login session dies for whatever reason (logout with
Richard> suspended job, network failure, etc.), the emacs will
Richard> remain, and someone else can connect to the machine and
Richard> have control of your emacs (sort of, since there is a
Richard> competing login process on the same pty).
Two points. First, there's no sort of about it: I've gotten
very good at taking over these Emacs processes and making use of the
shells to close down sessions. Second, Emacs is a good example, but
it isn't the only program that demonstrates the failure mpode.
Second, I've already sent in a telnetd patch to rel-eng. It
has already been applied to the telnetd in the telnet locker and to
the v5 source tree (where it really doesn't matter anyway). I suspect
that at least v5rlogind is also potentially effected, but it's pty
handling is good enough that it'll never grab an in-use pty, although
it still should try revoking them.
I believe I already installed a patched telnetd on tardis; I thought I sent you mail about this. If not, the one out of the telnetd locker should do. Let me know if you are able to reproduce the bug with that version of telnetd.
--Sam
