Re: login shells

Re: login shells

[Donn Cave]

> I think rshd is broken.

How should it work?  Would it be OK with you, if rshd exec'd commands
with "-" prefixed to the shell name in argv[0], thus causing them to
be "logins"?  That will force people who exec stty etc. in .rcrc to
bracket that stuff with an interactivity test, but it's fairly
straightforward.

	Donn Cave, University Computing Services, University of Washington
	donn@cac.washington.edu


--
``The use of history as therapy means the corruption of history as history.''
	-- Arthur Schlesinger

Chet Ramey, Case Western Reserve University	Internet: chet@po.CWRU.Edu

Re: login shells

[Donn Cave]

|              "who am i" in its long format tells you from where you have
| logged onto a machine.

But not with rsh, which is where it's really needed.  There's no utmp entry.

| $ rsh wherever who am i
| donn     tty??   Aug  7 08:30

	Donn Cave, University Computing Services, University of Washington
	donn@cac.washington.edu

Re: login shells

[malte]

In current UNIX there is a mechanism like in 10th edition UNIX, although
it's hidden. "who am i" in its long format tells you from where you have
logged onto a machine.

I use this feature to set the DISPLAY env. variable in my X11 environment.
It works on SunOS 4.*, Ultrix 4*, AIX 3.1*, and A/UX 2.* and maybe others
I don't have access to.

One could easily include a few lines which check for that
(hostname != "entry_in_etc_utmp").

Malte		(malte@techfak.uni-bielefeld.de)

Re: login shells

[Donn Cave]

| I like the 10th Edition method of putting a special variable in the
| environment of the invoked shell (REMOTE= or REMOTEHOST=otherhost or
| something like that).

That seems like a good idea, regardless.

But in order to use it, the shell has to be prepared to source some file
when it sees this variable (and then presumably delete the variable).
For rc, that would be .rcrc? so the effect would be similar to forcing
a "login" shell?

	Donn Cave, University Computing Services, University of Washington
	donn@cac.washington.edu

Re: login shells

[Chet Ramey]

> > I think rshd is broken.
> 
> How should it work?  Would it be OK with you, if rshd exec'd commands
> with "-" prefixed to the shell name in argv[0], thus causing them to
> be "logins"?  That will force people who exec stty etc. in .rcrc to
> bracket that stuff with an interactivity test, but it's fairly
> straightforward.

I like the 10th Edition method of putting a special variable in the
environment of the invoked shell (REMOTE= or REMOTEHOST=otherhost or
something like that).

Chet

--
``Someday soon, I'm gonna tell the world about the crying game...''

Chet Ramey			  Internet: chet@po.CWRU.Edu
Case Western Reserve University	  NeXT Mail: chet@macbeth.INS.CWRU.Edu

Re: login shells

[Boyd Roberts]

I think rshd is broken.

One of the things I like about rc is the absence of superfluous command
line options, no weird ${foo+_)()*&^$%&%=blah} and no obscure semantics.

I'd hate to see:

    $home/.rcrc is sourced it the first character in argv[0] is '-'
    or the -l' options is used, or fd 5 is a named pipe (on Wednesdays),
    or TIOCAREYOUANXTERM works and your window manager supports a predictive
    keyboard focus algorithm (with XAnothorFoulLibraryCall(3))...

login shells

[Donn Cave]

Speaking of login shells, have we discussed .rcrc vs. rsh yet?  I haven't
tried this with the latest version, but I believe that rc invoked by rshd
doesn't read its startup.  That means that if a program isn't in the machine's
standard execution path, it can't be invoked by name via rsh, without the
assistance of a wrapper script that sets path.  For example.

I'm not saying that it's impossible to cope with, or even hard for an
intermediate shell hacker.  Bash gets it right, that's all.

	Donn Cave, University Computing Services, University of Washington
	donn@cac.washington.edu