problems with scp

[Rachit Siamwalla]

I have an openssh server running on solaris 7 and the client running on
a clean RH linux 6.1 box. I am using the latest release (2.1.1p1 -- i
had the known solaris utmp problem and the read_key problem when using
2.1.0p2 so i upgraded).

scp from the linux machine to solaris machine gives a "lost connection"
message

[root at dog /root]# scp foo.bar solmachine:/tmp/
lost connection

the file foo.bar does not even have to exist for this to happen.

The -v and the -d flags don't seem to give more information (i compared
the output with a linux-linux scp).

strace / truss information can be made available, although it doesn't
seem like any system call failed to cause this to happen. Looking from
the source code, it looks like the client scp does a read (using the
atomicio call), expects the message to be a certain length, but it
isn't. So it assumes the server disconnected (which seems to be a valid
assumption), and prints "lost connection" then disconnects. However
looking at the server side, it thought the client disconnected. Anyways,
did anyone else ever encounter this problem (so i can confirm that it
isn't my hallucination)?

(i'll probably go back to using datafellows ssh).

BTW, i am not a member of this list, so please cc any responses to me.

thanx.

-rchit