[9fans] ssh patch

[9fans] ssh patch

[Richard C Bilson]

I've been having some trouble when piping files to unix systems via
ssh.  I was finding that a number of EOT characters (^D) were being
appended to the file.  Investigating, I discovered that ssh.c can send
multiple eof messages at end-of-file.  I don't know enough about the
protocol to know if this behavior should be allowed, but avoiding the
excess messages fixes my problem.

I don't know what the preferred form for patches is -- I was expecting to
use 'diff -c', but that's not happening, so here's a straight diff:

term% diff -n /sys/src/cmd/ssh/ssh.c ssh.c
/sys/src/cmd/ssh/ssh.c:548 a ssh.c:549,550
> 			else
> 				continue;

(this is in function "fromstdin", just after the break after the test
"++eof > 32")

- Richard

Re: [9fans] ssh patch

[Russ Cox]

There's one problem with this patch, namely that
it breaks cooked interactive mode (ssh -C).
In cooked interactive mode, a zero-length read
needs to turn into a ^D on the wire.

I changed the code to send the ^D only in
interactive mode.  In non-interactive mode
(e.g., when you're piping something into the
ssh program), you get the immediate eof that
your original patch was after.

I think this should solve your problem.
Let me know if it doesn't.

Russ

Re: [9fans] ssh patch

[Mike Fletcher]

Speaking of ssh, (when) can we expect an ssh version 2
client for Plan 9 ??  All of our unix boxes expect
version 2 authentication now.

Any one working on it?

Thanks,

-Mike Fletcher
 Sys Admin EECS
 Tulane University

Re: [9fans] ssh patch

[William Josephson]

On Tue, Apr 15, 2003 at 02:53:02PM -0500, Mike Fletcher wrote:
> Speaking of ssh, (when) can we expect an ssh version 2
> client for Plan 9 ??  All of our unix boxes expect
> version 2 authentication now.
>
> Any one working on it?

Yes.  If all goes well there will be a preliminary
version in a couplefew weeks.  I started back in
August and mentioned in December or so that I had
the beginnings of a client (and eventually a server),
but there has been a 7+ month hiatus -- I only got
back to it last week when rsc pestered me about it
again.  I have a mostly working transport protocol
implementation, so if I don't get distracted there
ought to be something useful sometime in May.