Subject: Re: Command execution end

Re: Command execution end

From: vti <viacheslav.t_at_gmail.com>
Date: Tue, 4 Aug 2009 17:17:16 +0200

2009/8/4 vti <viacheslav.t_at_gmail.com>:
> 2009/8/4 Peter Stuge <peter_at_stuge.se>:
>> vti wrote:
>>> > RFC 4254 says to send SSH_MSG_CHANNEL_CLOSE. _channel_send_eof()
>>> > should be used only internally in libssh2. Maybe also marked that
>>> > way in the API somehow? Yes, might break ABI.
>>> >
>>> > Please try _channel_close() and _channel_wait_closed() instead.
>>>
>>> I tried all that before writing to the list =/ No success.
>>
>> So a few emails could have been saved if you had mentioned that. What
>> else have you tried?
>>
>> Could you provide a minimal program demonstrating the problem?
>>
>>
>>> while (libssh2_channel_close(channel) == LIBSSH2_ERROR_EAGAIN);
>>> while (libssh2_channel_wait_closed(channel) == LIBSSH2_ERROR_EAGAIN);
>>
>> Ok. Do you know if at least _close() returns quickly?
>
> There is no delay.
>
>>> From the manual page:
>>>        .. sending an SSH_MSG_CLOSE packet to the remote host ..
>>>
>>> Is SSH_MSG_CLOSE is the same as SSH_MSG_CHANNEL_CLOSE?
>>
>> Yes, that's an error in the man page.
>>
>> What happens on the server side? strace? sshd -ddd?
>>
>> And of course, try to enable libssh2 debugging.
>
> Meanwhile I will provide my minimal failing example and trace sshd,
> here is a trace
> log in attachment. I am trying to run '/bin/mkdir /tmp/foo'.

And attachment :)

_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel

Received on 2009-08-04