On Monday, June 12th 2017 14:23:38, Andreas Schneider

• changed task status from "Wontfix" to "New"

On Monday, June 12th 2017 14:23:16, Andreas Schneider

• changed task status from "New" to "Wontfix"

On Thursday, May 11th 2017 12:19:33, Patrick Fial wrote:

Actually I was able to pinpoint the "bad" remote command: the "npm install" command seems to request a PTY, which packet type 98 indicates. However, the channel is closed thereafter, probably because I requested a shell on the channel, but NOT a PTY. So either the server disconnects because libssh denies the pty request, or probably libssh closes the channel because I did not open a PTY?
I was kind of able to confirm this assumption by changing my program code to also request a PTY via libssh commands, but this is not a solution, because parsing the resulting command/channel output in shell/pty mode will be cumbersome/impossible.

Either way, I wonder if there is a solution for this issue. Would this mean that it is not possible to execute remote commands in non-pty mode which might request a PTY?

I tried to use the channelptyrequest//function, like so:

		struct ssh//channel//callbacks//struct cb = {
			.userdata = NULL,
			.channel//pty//request//function = onPtyRequest
		};

		ssh//callbacks//init(&cb);
		ssh//set//channel//callbacks(channel, &cb);
[...]
       int onPtyRequest(ssh//session session, ssh//channel channel, const char **term, int width, int height, int pxwidth, int pwheight, void **userdata)
   	{
   		printf("######################### PTY request received ##############\n");
   		return -1;
   	}

My idea was that I could return -1 to deny the pty request, but this only makes the program crash with signal 11.

** thread #1, queue = 'com.apple.main-thread', stop reason = EXC//BAD//ACCESS (code=EXC//I386//GPFLT)
  ** frame #0: 0x0000000100001cef a.out
    frame #1: 0x00000001000c6da1 libssh.4.dylib`ssh//socket//pollcallback + 1857
    frame #2: 0x00000001000c1e09 libssh.4.dylib`ssh//poll//ctx//dopoll + 441
    frame #3: 0x00000001000c3af3 libssh.4.dylib`ssh//handle//packets + 291
    frame #4: 0x00000001000c37c8 libssh.4.dylib`ssh//handle//packets//termination + 232
    frame #5: 0x000000010009f9dc libssh.4.dylib`ssh//channel//read//timeout + 396
    frame #6: 0x000000010009f83f libssh.4.dylib`ssh//channel//read + 47
    frame #7: 0x000000010000142b a.out`readBytes(aBuffer=0x00000001003070f0, fromStderr=0, until="9903373aee394f16a077720cc5f58e80") at main2.cc:303
    frame #8: 0x000000010000112a a.out`execute(command="date") at main2.cc:241
    frame #9: 0x0000000100000b5b a.out`main(argc=1, argv=0x00007fff5fbff668) at main2.cc:75
    frame #10: 0x00007fffaacc9235 libdyld.dylib`start + 1
    frame #11: 0x00007fffaacc9235 libdyld.dylib`start + 1

On Monday, April 24th 2017 13:07:12, Andreas Schneider

• set category to "misc"
• assigned the task to Andreas Schneider

On Monday, April 24th 2017 13:06:51, Andreas Schneider wrote:

The patch is ok. Some questions:

Shouldn't we add an option to compile the binary (-DWITH//FUZZING=ON)?

Is libFuzzer part of clang or a project on its own?

On Thursday, June 8th 2017 22:14:21, Antenore Gatta wrote:

Hi Aris,

In Remmina they have opened a bug probably related to this.

https://github.com/FreeRDP/Remmina/issues/1187

Do you think it's the same case? Otherwise do you know how can I deal wit multiple keys using sshuserauthpublickey//auto ?

Thank you!!!

On Sunday, March 19th 2017 12:34:13, Alberto Delgado wrote:

Same here. Tested with last version of Remmina.
I suspect that it close TCP socket when a packet loss occurs. Very often on wireless networks.
Ubuntu's ssh command its very much robust and reestablish session transparently.
I tested unplugging wire ethernet from my wireless access point and plugging again (no wireless link lost on laptop). SSH command reenable session even if I type some words during AP wire disconnection or show all remaining 'top refresh' together when I reconnect wire to AP. LibSSH (Remmina) close window as soon I unplug AP's ethernet wire.
It can explain why libssh connection over openssh tunnel doesnt close. Openssh recover any packet lost with very low delay.
I wish this info could be usefull.
Regards

On Wednesday, October 12th 2016 08:20:10, Andreas Schneider

• changed task status from "Feedback" to "Assigned"

On Monday, October 10th 2016 04:42:16, Alfred Yang wrote:

I confirm too.

Client (OS name and version): Ubuntu 16.04.1
Remmina version (remmina --version): 1.2.0-rcgit-14 (git n/a)
Desktop environment (GNOME, Unity, KDE, ..): Unity
Connecting to (OS and version): CentOS 6.8
Connecting via (RDP, VNC, ...): VNC

below is a log for my VNC over SSH tunnel; It also happens on RDP over SSH tunnel

[SSH] ssh//config//parse//file: Reading configuration data from /home/<masked>/.ssh/config
[SSH] ssh//config//parse//line: Unsupported option: GSSAPIAuthentication, line: 2

[SSH] ssh//config//parse//line: Unsupported option: GSSAPIKeyExchange, line: 4

[SSH] ssh//config//parse//line: Unsupported option: GSSAPITrustDNS, line: 5

[SSH] ssh//config//parse//line: Unsupported option: ForwardX11Trusted, line: 10

[SSH] ssh//config//parse//line: Unsupported option: PubkeyAuthentication, line: 11

[SSH] ssh//config//parse//line: Unsupported option: ForwardX11Trusted, line: 16

[SSH] ssh//config//parse//line: Unsupported option: PubkeyAuthentication, line: 17

[SSH] ssh//config//parse//line: Unsupported option: Match, line: 20

[SSH] ssh//config//parse//line: Unsupported option: ForwardX11Trusted, line: 22

[SSH] ssh//config//parse//line: Unsupported option: PubkeyAuthentication, line: 23

[SSH] ssh//config//parse//file: Reading configuration data from /etc/ssh/ssh//config
[SSH] ssh//config//parse//line: Unsupported option: SendEnv, line: 53

[SSH] ssh//config//parse//line: Unsupported option: HashKnownHosts, line: 54

[SSH] ssh//config//parse//line: Unsupported option: GSSAPIAuthentication, line: 55

[SSH] ssh//connect: libssh 0.6.3 (c) 2003-2014 Aris Adamantiadis, Andreas Schneider, and libssh contributors. Distributed under the LGPL, please refer to COPYING file for information about your rights, using threading threads//pthread
[SSH] getai: host <masked> matches an IP address
[SSH] ssh//socket//connect: Nonblocking connection socket: 16
[SSH] ssh//connect: Socket connecting, now waiting for the callbacks to work
[SSH] ssh//connect: ssh//connect: Actual timeout : 10000
[SSH] ssh//socket//pollcallback: Received POLLOUT in connecting state
[SSH] socket//callback//connected: Socket connection callback: 1 (0)
[SSH] callback//receive//banner: Received banner: SSH-1.99-OpenSSH//5.3
[SSH] ssh//client//connection//callback: SSH server banner: SSH-1.99-OpenSSH//5.3
[SSH] ssh//analyze//banner: Analyzing banner: SSH-1.99-OpenSSH//5.3
[SSH] ssh//analyze//banner: We are talking to an OpenSSH client version: 5.3 (50300)
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] ssh//packet//socket//callback: packet: read type 20 [len=836,padding=10,comp=825,payload=825]
[SSH] ssh//packet//process: Dispatching handler for packet type 20
[SSH] ssh//list//kex: kex algos: diffie-hellman-group-exchange-sha256,diffie-hellman-group-exchange-sha1,diffie-hellman-group14-sha1,diffie-hellman-group1-sha1
[SSH] ssh//list//kex: server host key algo: ssh-rsa,ssh-dss
[SSH] ssh//list//kex: encryption client->server: aes128-ctr,aes192-ctr,aes256-ctr,arcfour256,arcfour128,aes128-cbc,3des-cbc,blowfish-cbc,cast128-cbc,aes192-cbc,aes256-cbc,arcfour,rijndael-cbc`lysator.liu.se
[SSH] ssh//list//kex: encryption server->client: aes128-ctr,aes192-ctr,aes256-ctr,arcfour256,arcfour128,aes128-cbc,3des-cbc,blowfish-cbc,cast128-cbc,aes192-cbc,aes256-cbc,arcfour,rijndael-cbc`lysator.liu.se
[SSH] ssh//list//kex: mac algo client->server: hmac-md5,hmac-sha1,umac-64`openssh.com,hmac-sha2-256,hmac-sha2-512,hmac-ripemd160,hmac-ripemd160`openssh.com,hmac-sha1-96,hmac-md5-96
[SSH] ssh//list//kex: mac algo server->client: hmac-md5,hmac-sha1,umac-64`openssh.com,hmac-sha2-256,hmac-sha2-512,hmac-ripemd160,hmac-ripemd160`openssh.com,hmac-sha1-96,hmac-md5-96
[SSH] ssh//list//kex: compression algo client->server: none,zlib`openssh.com
[SSH] ssh//list//kex: compression algo server->client: none,zlib`openssh.com
[SSH] ssh//list//kex: languages client->server: 
[SSH] ssh//list//kex: languages server->client: 
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=1
[SSH] ssh//knownhosts//algorithms: server <masked>:22 has ssh-rsa in known//hosts
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] ssh//client//select//hostkeys: Changing host key method to "ssh-rsa"
[SSH] ssh//list//kex: kex algos: curve25519-sha256`libssh.org,ecdh-sha2-nistp256,diffie-hellman-group14-sha1,diffie-hellman-group1-sha1
[SSH] ssh//list//kex: server host key algo: ssh-rsa
[SSH] ssh//list//kex: encryption client->server: aes256-ctr,aes192-ctr,aes128-ctr,aes256-cbc,aes192-cbc,aes128-cbc,blowfish-cbc,3des-cbc,des-cbc-ssh1
[SSH] ssh//list//kex: encryption server->client: aes256-ctr,aes192-ctr,aes128-ctr,aes256-cbc,aes192-cbc,aes128-cbc,blowfish-cbc,3des-cbc,des-cbc-ssh1
[SSH] ssh//list//kex: mac algo client->server: hmac-sha1
[SSH] ssh//list//kex: mac algo server->client: hmac-sha1
[SSH] ssh//list//kex: compression algo client->server: none
[SSH] ssh//list//kex: compression algo server->client: none
[SSH] ssh//list//kex: languages client->server: 
[SSH] ssh//list//kex: languages server->client: 
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] packet//send2: packet: wrote [len=404,padding=6,comp=397,payload=397]
[SSH] packet//send2: packet: wrote [len=268,padding=5,comp=262,payload=262]
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] ssh//packet//socket//callback: packet: read type 31 [len=828,padding=9,comp=818,payload=818]
[SSH] ssh//packet//process: Dispatching handler for packet type 31
[SSH] ssh//packet//dh//reply: Received SSH//KEXDH//REPLY
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] packet//send2: packet: wrote [len=12,padding=10,comp=1,payload=1]
[SSH] ssh//client//dh//reply: SSH//MSG//NEWKEYS sent
[SSH] ssh//packet//socket//callback: Processing 16 bytes left in socket buffer
[SSH] ssh//packet//socket//callback: packet: read type 21 [len=12,padding=10,comp=1,payload=1]
[SSH] ssh//packet//process: Dispatching handler for packet type 21
[SSH] ssh//packet//newkeys: Received SSH//MSG//NEWKEYS
[SSH] crypt//set//algorithms2: Set output algorithm to aes256-ctr
[SSH] crypt//set//algorithms2: Set input algorithm to aes256-ctr
[SSH] ssh//pki//signature//verify//blob: Going to verify a ssh-rsa type signature
[SSH] ssh//packet//newkeys: Signature verified and valid
[SSH] ssh//connect: ssh//connect: Actual state : 7
[SSH] packet//send2: packet: wrote [len=28,padding=10,comp=17,payload=17]
[SSH] ssh//service//request: Sent SSH//MSG//SERVICE//REQUEST (service ssh-userauth)
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] ssh//packet//socket//callback: packet: read type 6 [len=28,padding=10,comp=17,payload=17]
[SSH] ssh//packet//process: Dispatching handler for packet type 6
[SSH] ssh//packet//service//accept: Received SSH//MSG//SERVICE//ACCEPT
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] packet//send2: packet: wrote [len=44,padding=5,comp=38,payload=38]
[SSH] ssh//packet//socket//callback: packet: read type 51 [len=44,padding=19,comp=24,payload=24]
[SSH] ssh//packet//process: Dispatching handler for packet type 51
[SSH] ssh//packet//userauth//failure: Access denied. Authentication that can continue: publickey,password
[SSH] ssh//packet//userauth//failure: Access denied. Authentication that can continue: publickey,password
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=1
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=0
[SSH] match//hashed//host: Matching a hashed host: [<masked>]:22 match=0
[SSH] match//hashed//host: Matching a hashed host: <masked> match=1
[SSH] agent//talk: Request length: 1
[SSH] agent//talk: Response length: 322
[SSH] ssh//agent//get//ident//count: Answer type: 12, expected answer: 12
[SSH] ssh//agent//get//ident//count: Agent count: 1
[SSH] ssh//userauth//agent: Trying identity <masked>
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] packet//send2: packet: wrote [len=348,padding=11,comp=336,payload=336]
[SSH] ssh//packet//socket//callback: packet: read type 60 [len=300,padding=6,comp=293,payload=293]
[SSH] ssh//packet//process: Dispatching handler for packet type 60
[SSH] ssh//packet//userauth//pk//ok: Received SSH//USERAUTH//PK//OK/INFO//REQUEST/GSSAPI//RESPONSE
[SSH] ssh//packet//userauth//pk//ok: Assuming SSH//USERAUTH//PK//OK
[SSH] ssh//userauth//agent: Public key of <masked> accepted by server
[SSH] agent//talk: Request length: 650
[SSH] agent//talk: Response length: 276
[SSH] ssh//socket//unbuffered//write: Enabling POLLOUT for socket
[SSH] packet//send2: packet: wrote [len=620,padding=8,comp=611,payload=611]
[SSH] ssh//packet//socket//callback: packet: read type 52 [len=12,padding=10,comp=1,payload=1]
[SSH] ssh//packet//process: Dispatching handler for packet type 52
[SSH] ssh//packet//userauth//success: Authentication successful
[SSH] ssh//packet//userauth//success: Received SSH//USERAUTH//SUCCESS
[VNC]VNC server supports protocol version 3.8 (viewer 3.8)
[VNC]We have 1 security types to read
[VNC]0) Received security type 2
[VNC]Selecting security type 2 (0/1 in the list)
[VNC]Selected Security Scheme 2
[glibsecret] found password for file /home/<masked>/.remmina/1422409939077.remmina
[VNC]VNC authentication succeeded
[VNC]Desktop name "<masked>:247 (<masked>)"
[VNC]Connected to VNC server, using protocol version 3.8
[VNC]VNC server default format:
[VNC]  16 bits per pixel.
[VNC]  Least significant byte first in each pixel.
[VNC]  TRUE colour: max red 31 green 63 blue 31, shift red 11 green 5 blue 0
[SSH] tunnel has been disconnected. Reason: ssh//channel//poll() returned an error : Socket error: Connection timed out
[VNC]VNC server closed connection

From the log, it tells "Connection timed out", but at the same time, my ssh terminal session is still alive and working(through SecureCRT), so my network is just fine.

On Saturday, October 8th 2016 16:43:51, Andreas Schneider wrote:

Waiting on feedback

and:

• changed task status from "New" to "Feedback"

On Friday, September 9th 2016 11:44:24, Aris Adamantiadis wrote:

Unfortunately we're a little short of information here to reproduce the problem. If I understand well:

• It happens on many different server types (Ubuntu, debian, centos, windows)
• It happens "randomly" (I need more precision here. Does it happen after 5 minutes, when you leave the screen idling away, after a big transfer, ...?)
• Connecting via: SSH => does it mean it's pure SSH or VNC over SSH? Does it happen with RDP over SSH too?
• The log output seems truncated after authentication. If there's a way to generate a complete log dump (it could easily take a few megabytes) it would greatly help, especially the last hundred lines

Thanks,

Aris

On Wednesday, August 31st 2016 20:53:03, Antenore Gatta wrote:

I confirm, it happens from time to time.

An other user reported the same on Kubuntu:

Client (OS name and version): Kubuntu 14.04.5
Remmina version (remmina --version): 1.2.0-rcgit-15
Desktop environment (GNOME, Unity, KDE, ..): KDE 4.14
Connecting to (OS and version): Ubuntu, Debian, Centos...
Connecting via (RDP, VNC, ...): SSH

On Tuesday, August 23rd 2016 12:25:11, Milian Reichardt wrote:

UPDATE:
My current setup looks like this:
The Windows Machine is a Terminal Server I don't need RDP so I installed Cygwin connected to it over SSH and then connected from that terminal Server to our production systems.
The Connection to the Terminal Server is getting disconnected.

WORKAROUND:
I establish an SSH Tunnel to the Terminal Server via Gnome-terminal (which uses OpenSSH, here is the connection completely stable, tested it for 3 Hours) and then use Remmina and libssh to connect through the tunnel to the production environment.
But I still can't tell where exactly the problem is.

On Wednesday, November 23rd 2016 12:34:18, David Wedderwille wrote:

Added Signed-off-by

and:

On Tuesday, August 9th 2016 07:28:18, David Wedderwille wrote:

I've implemented the function sshpkiexportprivkeybase64 which is compatible to sshpkiimportprivkeybase64.

and:

On Monday, June 6th 2016 09:20:46, Andreas Schneider wrote:

Aris is currently working on that ...

and:

• assigned the task to Aris Adamantiadis

On Wednesday, May 4th 2016 13:00:49, Andreas Schneider wrote:

a) Against which SSH server are you using this?
b) Can you write an example or a test case for our test infrastructure?

On Sunday, March 13th 2016 09:44:25, a b wrote:

it seems that the strings terminate with an SPACE instead of \0

On Monday, February 22nd 2016 21:12:54, Gaetan de Villele wrote:

Thank you for the reply.

I wasn't talking about "subsystems" nor "requests we send inside channels", (5.4. "Channel-Specific Requests" in RFC 4254)
I probably should have put more details in my question.

I was talking about the type of a channel itself. (5.1. "Opening a Channel" in RFC 4254)
For exemple, for sftp, we have to open a channel of type "session" and then send a request inside this channel to ask the server if it supports the "sftp subsystem"...
I would like libssh to allow for channels that are not of type "session" but of an arbitrarily type.

in channels.c the non-exposed function static int channel//open(ssh//channel channel, const char **type, int window, int maxpacket, ssh//buffer payload) is called multiple times with hardcoded channel types like "session", "direct-tcpip", "forwarded-tcpip", "x11" ...
Basically I am suggesting writing a function
int ssh//channel//open//custom(ssh//channel channel, ssh//string type, ssh//buffer payload)
in addition to the existing
int ssh//channel//open//session(ssh//channel channel)
int ssh//channel//open//auth//agent(ssh//channel channel)
int ssh//channel//open//forward(ssh//channel channel, const char **remotehost, int remoteport, const char **sourcehost, int local port)
...

Feel free to ask me any question if it is still unclear.
Thank you.

On Monday, February 22nd 2016 07:03:29, Andreas Schneider wrote:

This is already supported, see

http://api.libssh.org/master/group////libssh////channel.html#ga06024b070f9b2a3d6964b79ae36695b7

and the RFCs.

On Tuesday, October 20th 2015 18:31:29, Weiyin He wrote:

Btw, this patch has worked well for me: http://www.libssh.org/archive/libssh/2014-03/0000023.html

On Tuesday, October 20th 2015 18:13:22, Andreas Schneider

• changed task status from "New" to "Assigned"
• assigned the task to Aris Adamantiadis

On Tuesday, October 20th 2015 18:20:25, Andreas Schneider wrote:

Call sshinit() before you do any stuff with SSH and sshfinalize() in the end...

On Monday, May 11th 2015 11:39:59, Yuki Yuki wrote:

Yuki Yuki wrote:

Yuki Yuki wrote:

Andreas Schneider wrote:

Run with valgrind to find out what is wrong ...

I attached small project with problem

I did log valgrind.

and:

On Monday, May 11th 2015 11:14:49, Yuki Yuki wrote:

Yuki Yuki wrote:

Andreas Schneider wrote:

Run with valgrind to find out what is wrong ...

I attached small project with problem

and:

On Monday, May 11th 2015 10:09:52, Yuki Yuki wrote:

Andreas Schneider wrote:

Run with valgrind to find out what is wrong ...

i can’t do it, because Valgrind can’t support my osx.

checking for the kernel version... unsupported (14.3.0)
configure: error: Valgrind works on Darwin 10.x, 11.x, 12.x and 13.x (Mac OS X 10.6/7/8/9)

I updated to libssh-0.7.0, i have same problems

(22118,0x103e07000) malloc: ** error for object 0x7fde23406860: pointer being freed was not allocated

• set a breakpoint in mallocerrorbreak to debug

or

(22213,0x10028a000) malloc: ** error for object 0x1003058b8: incorrect checksum for freed object - object was probably modified after being freed.

• set a breakpoint in mallocerrorbreak to debug

bt example

• thread #2: tid = 0x4ca7f2, 0x00007fff97603282 libsystemmalloc.dylib`tinyfreelistaddptr + 108, stop reason = EXCBAD//ACCESS (code=1, address=0x100600c9e)
• frame #0: 0x00007fff97603282 libsystemmalloc.dylib`tinyfreelistaddptr + 108 frame #1: 0x00007fff9760053e libsystemmalloc.dylib`tinymallocfromfreelist + 1450 frame #2: 0x00007fff975fef50 libsystemmalloc.dylib`szonemallocshouldclear + 317 frame #3: 0x00007fff975fe877 libsystemmalloc.dylib`malloczonemalloc + 71 frame #4: 0x00007fff975fd395 libsystemmalloc.dylib`malloc + 42 frame #5: 0x0000000100026259 test1`sshbufferunpackva + 937 frame #6: 0x0000000100026676 test1`sshbufferunpack + 134 frame #7: 0x0000000100022883 test1`sshpacketuserauthfailure + 67 frame #8: 0x0000000100039e0b test1`sshpacketprocess + 187 frame #9: 0x0000000100039c69 test1`sshpacketsocketcallback + 1561 frame #10: 0x000000010004122b test1`sshsocketpollcallback + 811 frame #11: 0x000000010003f7b7 test1`sshpollctxdopoll + 231 frame #12: 0x0000000100040a2f test1`sshhandlepackets + 191 frame #13: 0x000000010004086b test1`sshhandlepacketstermination + 155 frame #14: 0x0000000100024147 test1`sshuserauthpassword + 359

On Monday, May 11th 2015 08:55:07, Andreas Schneider wrote:

Run with valgrind to find out what is wrong ...

On Tuesday, October 20th 2015 18:21:21, Andreas Schneider

• assigned the task to Aris Adamantiadis

On Tuesday, October 20th 2015 18:21:54, Andreas Schneider

• assigned the task to Andreas Schneider

On Friday, January 1st 2016 21:18:55, quentin garnier wrote:

I get the result of commands with sshchannelread and sshchannelselect. it works well.

I have tried to use events and i have problems:

• channelexitstatus//function seems to not be launched
• channeldatafunction can be launched or not
• i have segmentation fault

My process:

• i connect a session
• open channels: CA and CB
• create event
• add my session in the event
• set callback for channel CA
• execute commands in Channels CA and CB (only the execution with sshchannelrequest//exec)
• open two other channels CC and CD
• call ssheventdopoll to launch callbacks

I have done something wrong ? Maybe you need the trace with verbosity max ? Thanks for the help.

On Thursday, December 31st 2015 13:00:30, quentin garnier wrote:

Ok, i have find a solution for the channel id: i use localchannel integer and remotechannel integer from the struct channel. Well better than the pointer address.

On Monday, December 28th 2015 14:59:26, quentin garnier wrote:

I have a question. I would like to use following functions:

• sshchannelread//timeout
• sshchannelselect

Like that i could launch X channels with X commands in parrallel. But i don't know how to identify the channel.
I don't really want to use pointer value. Not very elegant. If i can have a function to get the stdout fd of a channel.

On Monday, December 28th 2015 12:09:29, quentin garnier wrote:

I have beginning to work on perl binding here:
https://github.com/garnier-quentin/perl-libssh

You're writing in doc:

Warning: sshselect() and sshchannel//select() are not relevant anymore,
since libssh is about to provide an easier system for asynchronous
communications. This subsection should be removed then. **

What we should use instead ?

On Wednesday, December 17th 2014 09:34:53, Anonymous wrote:

I'd like to use ftp instead. But i didn't find a timeout option if nothing is sent (like 'ServerAliveInterval' and 'ServerAliveCountMax' options).

On Wednesday, December 17th 2014 09:29:33, Andreas Schneider wrote:

Why don't you use sftp instead of this horrible scp protocol?

On Friday, January 16th 2015 21:41:54, Michael DePaulo wrote:

Andreas Schneider wrote:

With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time.

I did some research :

1. It appears that there is no actively maintained port of OpenSSH (and therefore ssh-agent) to Windows. In order to not be affected by a vulnerability, one needs to use OpenSSH 6.7p1 or later.

2. In the past, NoMachine ported SSH to Windows. But they are not maintaining it anymore. The last ported version was 5.9p1.

https://www.nomachine.com/AR11K00739
https://www.nomachine.com/NoMachine-OSS-ports

3. There is actively maintained software such as "OpenSSH for Windows", which is at 6.7p1. But they are actually OpenSSH for Cygwin with an easy Windows installer.

http://www.mls-software.com/opensshd.html

If you'd like, I can test cygwin ssh-agent 6.7p1 with native Windows libssh and see if it actually works. But I do not feel it is appropriate to maintain support for native Windows ssh-agent so long as it is unmaintained.

On Sunday, January 11th 2015 14:27:05, Michael DePaulo wrote:

Back in September, my friend Michael Frederick and I (gmail: psududemike) managed to port that patch to 0.6.3. We've been using it successfully in X2Go Client for Windows 4.0.3.0 and later.
That patch still applies to 0.6.4. I've attached the patch.

Comments are welcome.

I just signed my DCO. Mike Frederick assigned copyright of this patch to me, but if need be, he can sign the DCO also.

However, I am not sure about the DCO from the authors for the original patch. Actually, there were 2 different versions of the original patch:
0001-implement-support-for-putty-s-pageant.patch - which was superseded by:
0001-implement-support-for-putty-s-pageant-0.5.3.patch
And to make things more confusing, there were multiple versions of the 0.5.3 patch.
That ohloh.net link no longer works. But both original patches are available here.
https://projects.kde.org/projects/kdesupport/emerge/repository/revisions/master/show/portage/win32libs/libssh

The 2 authors of those patches appear to be:
Patrick Spendrin <ps//ml`gmx.de>
Patrick von Reth

So at this point, I am not sure how to proceed. Let me ask: Did those 2 authors ever submit the DCO?

and:

On Thursday, January 15th 2015 09:03:58, Andreas Schneider wrote:

Thanks for your work. Copyright doesn't need to be assigned if all people are individuals. The Certificate of Origin is only for people who send patches with Corporate Copyright!

The patch is far from being in a state that I would include it. We need to design a new API for this, similar to pki.c. With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time. So we need a way to build with both and register the backend at the agent engine. We can discuss this on IRC.

On Friday, April 11th 2014 00:07:08, Jon Simons wrote:

I've also written code as described here with the same movitation:

rc = ssh//event//dopoll(event, timeout//msec);

....

const char **lib//err//msg = ssh//get//error(session);

....

if (rc == SSH//ERROR) {
    /** XXX no better way to extract failure reason **/
    #define DISCONN//MSG "Received SSH//MSG//DISCONNECT: 11"
    int graceful//disconnect = (strstr(lib//err//msg, DISCONN//MSG) != NULL);

...

An extended error getter would be useful.

On Thursday, April 10th 2014 10:49:24, Stef Walter wrote:

The work around I'm using is screen-scraping for certain non-translated strings in sshgeterror() message.

On Tuesday, December 24th 2013 02:28:08, Hisashi Kojima wrote:

Thank you for accepting my suggestion. I read and understood the fixed code. I can not immediately upgrade my code base from 0.5.5 to 0.6.0rc3, but will try the new version. Thanks.

On Saturday, December 21st 2013 10:50:40, Andreas Schneider wrote:

This should be fixed in libssh 0.6.0rc3. Could you please try?

On Tuesday, July 17th 2012 15:59:50, Andreas Schneider wrote:

Can you please be more precise or add a code example?

and:

• set category to "session"

On Monday, July 15th 2013 21:01:42, Aris Adamantiadis

• changed the subject from "lib_ssh with Dolphin, not the same speed upload/download with sftp" to "Support ftp async upload (was w/ Dolphin, not the same speed upload/download with sftp)"
• unassigned Did G
• set target version to "0.7.0"

On Monday, November 28th 2011 20:19:26, Vineeth Pillai wrote:

patch added

and:

On Tuesday, August 9th 2011 19:53:04, Felix Hupfeld wrote:

when cutting writes into blocks of sshchannelwindow//size(channel) size, the writes work, but read still doesn't.

[2] Authentication successful
[2] Creating a channel 43 with 64000 window and 32768 max packet
[2] Received a CHANNELOPENCONFIRMATION for channel 43:43
[2] Remote window : 32000, maxpacket : 35000
[2] Channel request subsystem success
[2] Read (16384) buffered : 0 bytes. Window: 32000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 32000, writing 32000
[1] channelwrite wrote 32000 bytes
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] Channel receiving 32000 bytes data in 0 (local win=32000 remote win=64000)
[1] placing 32000 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=0 remote win=64000)
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] growing window (channel 43:43) to 128000 bytes
[2] Read (16384) buffered : 15616 bytes. Window: 128000
[2] Adding 128000 bytes to channel (43:43) (from 0 bytes)
E 10308 ssh.cpp:547 Short read: 15616 < 16384
I 10308 ssh.cpp:520 sshchannelwindowsize returns 128000, writing 68000
..\src\comm\ssh//test.cpp(13): error: Value of: conn->Read(buffer, 100000)

Actual: 0

Expected: 100000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 64000, writing 5
[1] channelwrite wrote 5 bytes
[1] channelwrite wrote 34990 bytes
[1] channelwrite wrote 33010 bytes
[2] Channel receiving 5 bytes data in 0 (local win=64000 remote win=60000)
[1] placing 5 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=63995 remote win=60000)
[2] Read (5) buffered : 5 bytes. Window: 63995
[2] growing window (channel 43:43) to 128000 bytes

and:

• unassigned Aris Adamantiadis

On Tuesday, September 20th 2011 08:19:39, Antonio Orefice wrote:

The same with 0.5.2 (1.1MBps)

On Saturday, September 17th 2011 06:44:31, Andreas Schneider wrote:

We will release 0.5.2 today. Please test and report back.

Moving bug to 0.5.3.

and:

• changed target version from "0.5.2" to "0.5.3"

On Friday, September 9th 2011 13:01:35, Antonio Orefice wrote:

0.5.1 gives the same speed as 0.5.0 under kio//sftp, i didn't tested 0.5.2; waiting for archlinux package.

On Thursday, September 8th 2011 17:11:54, Aris Adamantiadis wrote:

I developped a benchmark tool (tests/benchmark) which exposes the speed of sftp/scp/ssh. I do not see significant performance problems on 0.5.2, but maybe it's only affecting 0.5.0.

On Monday, August 29th 2011 09:03:18, Andreas Schneider

• Raised the task's priority to High
• changed target version from "0.5.1" to "0.5.2"

On Friday, July 22nd 2011 12:19:37, Aris Adamantiadis wrote:

Hi Antonio,

Thanks for you feedback. I will check ASAP what's causing this regression. We've made many architectural changes between 0.4 and 0.5 and there is room for such kind of problems.

Kr,

Aris

and:

• set category to "sftp"
• changed task status from "New" to "Assigned"
• assigned the task to Aris Adamantiadis
• Raised the task's priority to Normal
• set target version to "0.5.1"
• set estimated hours to 3

On Tuesday, October 20th 2015 18:21:21, Andreas Schneider

• assigned the task to Aris Adamantiadis

On Tuesday, October 20th 2015 18:21:54, Andreas Schneider

• assigned the task to Andreas Schneider

On Friday, January 1st 2016 21:18:55, quentin garnier wrote:

I get the result of commands with sshchannelread and sshchannelselect. it works well.

I have tried to use events and i have problems:

• channelexitstatus//function seems to not be launched
• channeldatafunction can be launched or not
• i have segmentation fault

My process:

• i connect a session
• open channels: CA and CB
• create event
• add my session in the event
• set callback for channel CA
• execute commands in Channels CA and CB (only the execution with sshchannelrequest//exec)
• open two other channels CC and CD
• call ssheventdopoll to launch callbacks

I have done something wrong ? Maybe you need the trace with verbosity max ? Thanks for the help.

On Thursday, December 31st 2015 13:00:30, quentin garnier wrote:

Ok, i have find a solution for the channel id: i use localchannel integer and remotechannel integer from the struct channel. Well better than the pointer address.

On Monday, December 28th 2015 14:59:26, quentin garnier wrote:

I have a question. I would like to use following functions:

• sshchannelread//timeout
• sshchannelselect

Like that i could launch X channels with X commands in parrallel. But i don't know how to identify the channel.
I don't really want to use pointer value. Not very elegant. If i can have a function to get the stdout fd of a channel.

On Monday, December 28th 2015 12:09:29, quentin garnier wrote:

I have beginning to work on perl binding here:
https://github.com/garnier-quentin/perl-libssh

You're writing in doc:

Warning: sshselect() and sshchannel//select() are not relevant anymore,
since libssh is about to provide an easier system for asynchronous
communications. This subsection should be removed then. **

What we should use instead ?

On Wednesday, December 17th 2014 09:34:53, Anonymous wrote:

I'd like to use ftp instead. But i didn't find a timeout option if nothing is sent (like 'ServerAliveInterval' and 'ServerAliveCountMax' options).

On Wednesday, December 17th 2014 09:29:33, Andreas Schneider wrote:

Why don't you use sftp instead of this horrible scp protocol?

On Friday, January 16th 2015 21:41:54, Michael DePaulo wrote:

Andreas Schneider wrote:

With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time.

I did some research :

1. It appears that there is no actively maintained port of OpenSSH (and therefore ssh-agent) to Windows. In order to not be affected by a vulnerability, one needs to use OpenSSH 6.7p1 or later.

2. In the past, NoMachine ported SSH to Windows. But they are not maintaining it anymore. The last ported version was 5.9p1.

https://www.nomachine.com/AR11K00739
https://www.nomachine.com/NoMachine-OSS-ports

3. There is actively maintained software such as "OpenSSH for Windows", which is at 6.7p1. But they are actually OpenSSH for Cygwin with an easy Windows installer.

http://www.mls-software.com/opensshd.html

If you'd like, I can test cygwin ssh-agent 6.7p1 with native Windows libssh and see if it actually works. But I do not feel it is appropriate to maintain support for native Windows ssh-agent so long as it is unmaintained.

On Sunday, January 11th 2015 14:27:05, Michael DePaulo wrote:

Back in September, my friend Michael Frederick and I (gmail: psududemike) managed to port that patch to 0.6.3. We've been using it successfully in X2Go Client for Windows 4.0.3.0 and later.
That patch still applies to 0.6.4. I've attached the patch.

Comments are welcome.

I just signed my DCO. Mike Frederick assigned copyright of this patch to me, but if need be, he can sign the DCO also.

However, I am not sure about the DCO from the authors for the original patch. Actually, there were 2 different versions of the original patch:
0001-implement-support-for-putty-s-pageant.patch - which was superseded by:
0001-implement-support-for-putty-s-pageant-0.5.3.patch
And to make things more confusing, there were multiple versions of the 0.5.3 patch.
That ohloh.net link no longer works. But both original patches are available here.
https://projects.kde.org/projects/kdesupport/emerge/repository/revisions/master/show/portage/win32libs/libssh

The 2 authors of those patches appear to be:
Patrick Spendrin <ps//ml`gmx.de>
Patrick von Reth

So at this point, I am not sure how to proceed. Let me ask: Did those 2 authors ever submit the DCO?

and:

On Thursday, January 15th 2015 09:03:58, Andreas Schneider wrote:

Thanks for your work. Copyright doesn't need to be assigned if all people are individuals. The Certificate of Origin is only for people who send patches with Corporate Copyright!

The patch is far from being in a state that I would include it. We need to design a new API for this, similar to pki.c. With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time. So we need a way to build with both and register the backend at the agent engine. We can discuss this on IRC.

On Friday, April 11th 2014 00:07:08, Jon Simons wrote:

I've also written code as described here with the same movitation:

rc = ssh//event//dopoll(event, timeout//msec);

....

const char **lib//err//msg = ssh//get//error(session);

....

if (rc == SSH//ERROR) {
    /** XXX no better way to extract failure reason **/
    #define DISCONN//MSG "Received SSH//MSG//DISCONNECT: 11"
    int graceful//disconnect = (strstr(lib//err//msg, DISCONN//MSG) != NULL);

...

An extended error getter would be useful.

On Thursday, April 10th 2014 10:49:24, Stef Walter wrote:

The work around I'm using is screen-scraping for certain non-translated strings in sshgeterror() message.

On Tuesday, December 24th 2013 02:28:08, Hisashi Kojima wrote:

Thank you for accepting my suggestion. I read and understood the fixed code. I can not immediately upgrade my code base from 0.5.5 to 0.6.0rc3, but will try the new version. Thanks.

On Saturday, December 21st 2013 10:50:40, Andreas Schneider wrote:

This should be fixed in libssh 0.6.0rc3. Could you please try?

On Monday, September 23rd 2013 14:48:15, Dustin Oprea wrote:

Sounds great.

On Monday, September 23rd 2013 13:56:59, Aris Adamantiadis wrote:

We could update the code of the button to issue a google search specific to the libssh website, what do you think ?

On Monday, September 9th 2013 14:58:37, Andreas Schneider wrote:

There is no code behind that button.

If you want to search our mailing list you can use:

http://news.gmane.org/gmane.network.ssh.libssh.general

On Tuesday, July 17th 2012 15:59:50, Andreas Schneider wrote:

Can you please be more precise or add a code example?

and:

• set category to "session"

On Monday, July 15th 2013 21:01:42, Aris Adamantiadis

• changed the subject from "lib_ssh with Dolphin, not the same speed upload/download with sftp" to "Support ftp async upload (was w/ Dolphin, not the same speed upload/download with sftp)"
• unassigned Did G
• set target version to "0.7.0"

On Monday, November 28th 2011 20:19:26, Vineeth Pillai wrote:

patch added

and:

On Tuesday, August 9th 2011 19:53:04, Felix Hupfeld wrote:

when cutting writes into blocks of sshchannelwindow//size(channel) size, the writes work, but read still doesn't.

[2] Authentication successful
[2] Creating a channel 43 with 64000 window and 32768 max packet
[2] Received a CHANNELOPENCONFIRMATION for channel 43:43
[2] Remote window : 32000, maxpacket : 35000
[2] Channel request subsystem success
[2] Read (16384) buffered : 0 bytes. Window: 32000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 32000, writing 32000
[1] channelwrite wrote 32000 bytes
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] Channel receiving 32000 bytes data in 0 (local win=32000 remote win=64000)
[1] placing 32000 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=0 remote win=64000)
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] growing window (channel 43:43) to 128000 bytes
[2] Read (16384) buffered : 15616 bytes. Window: 128000
[2] Adding 128000 bytes to channel (43:43) (from 0 bytes)
E 10308 ssh.cpp:547 Short read: 15616 < 16384
I 10308 ssh.cpp:520 sshchannelwindowsize returns 128000, writing 68000
..\src\comm\ssh//test.cpp(13): error: Value of: conn->Read(buffer, 100000)

Actual: 0

Expected: 100000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 64000, writing 5
[1] channelwrite wrote 5 bytes
[1] channelwrite wrote 34990 bytes
[1] channelwrite wrote 33010 bytes
[2] Channel receiving 5 bytes data in 0 (local win=64000 remote win=60000)
[1] placing 5 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=63995 remote win=60000)
[2] Read (5) buffered : 5 bytes. Window: 63995
[2] growing window (channel 43:43) to 128000 bytes

and:

• unassigned Aris Adamantiadis

On Tuesday, September 20th 2011 08:19:39, Antonio Orefice wrote:

The same with 0.5.2 (1.1MBps)

On Saturday, September 17th 2011 06:44:31, Andreas Schneider wrote:

We will release 0.5.2 today. Please test and report back.

Moving bug to 0.5.3.

and:

• changed target version from "0.5.2" to "0.5.3"

On Friday, September 9th 2011 13:01:35, Antonio Orefice wrote:

0.5.1 gives the same speed as 0.5.0 under kio//sftp, i didn't tested 0.5.2; waiting for archlinux package.

On Thursday, September 8th 2011 17:11:54, Aris Adamantiadis wrote:

I developped a benchmark tool (tests/benchmark) which exposes the speed of sftp/scp/ssh. I do not see significant performance problems on 0.5.2, but maybe it's only affecting 0.5.0.

On Monday, August 29th 2011 09:03:18, Andreas Schneider

• Raised the task's priority to High
• changed target version from "0.5.1" to "0.5.2"

On Friday, July 22nd 2011 12:19:37, Aris Adamantiadis wrote:

Hi Antonio,

Thanks for you feedback. I will check ASAP what's causing this regression. We've made many architectural changes between 0.4 and 0.5 and there is room for such kind of problems.

Kr,

Aris

and:

• set category to "sftp"
• changed task status from "New" to "Assigned"
• assigned the task to Aris Adamantiadis
• Raised the task's priority to Normal
• set target version to "0.5.1"
• set estimated hours to 3

On Tuesday, October 20th 2015 18:21:21, Andreas Schneider

• assigned the task to Aris Adamantiadis

On Tuesday, October 20th 2015 18:21:54, Andreas Schneider

• assigned the task to Andreas Schneider

On Friday, January 1st 2016 21:18:55, quentin garnier wrote:

I get the result of commands with sshchannelread and sshchannelselect. it works well.

I have tried to use events and i have problems:

• channelexitstatus//function seems to not be launched
• channeldatafunction can be launched or not
• i have segmentation fault

My process:

• i connect a session
• open channels: CA and CB
• create event
• add my session in the event
• set callback for channel CA
• execute commands in Channels CA and CB (only the execution with sshchannelrequest//exec)
• open two other channels CC and CD
• call ssheventdopoll to launch callbacks

I have done something wrong ? Maybe you need the trace with verbosity max ? Thanks for the help.

On Thursday, December 31st 2015 13:00:30, quentin garnier wrote:

Ok, i have find a solution for the channel id: i use localchannel integer and remotechannel integer from the struct channel. Well better than the pointer address.

On Monday, December 28th 2015 14:59:26, quentin garnier wrote:

I have a question. I would like to use following functions:

• sshchannelread//timeout
• sshchannelselect

Like that i could launch X channels with X commands in parrallel. But i don't know how to identify the channel.
I don't really want to use pointer value. Not very elegant. If i can have a function to get the stdout fd of a channel.

On Monday, December 28th 2015 12:09:29, quentin garnier wrote:

I have beginning to work on perl binding here:
https://github.com/garnier-quentin/perl-libssh

You're writing in doc:

Warning: sshselect() and sshchannel//select() are not relevant anymore,
since libssh is about to provide an easier system for asynchronous
communications. This subsection should be removed then. **

What we should use instead ?

On Wednesday, December 17th 2014 09:34:53, Anonymous wrote:

I'd like to use ftp instead. But i didn't find a timeout option if nothing is sent (like 'ServerAliveInterval' and 'ServerAliveCountMax' options).

On Wednesday, December 17th 2014 09:29:33, Andreas Schneider wrote:

Why don't you use sftp instead of this horrible scp protocol?

On Friday, January 16th 2015 21:41:54, Michael DePaulo wrote:

Andreas Schneider wrote:

With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time.

I did some research :

1. It appears that there is no actively maintained port of OpenSSH (and therefore ssh-agent) to Windows. In order to not be affected by a vulnerability, one needs to use OpenSSH 6.7p1 or later.

2. In the past, NoMachine ported SSH to Windows. But they are not maintaining it anymore. The last ported version was 5.9p1.

https://www.nomachine.com/AR11K00739
https://www.nomachine.com/NoMachine-OSS-ports

3. There is actively maintained software such as "OpenSSH for Windows", which is at 6.7p1. But they are actually OpenSSH for Cygwin with an easy Windows installer.

http://www.mls-software.com/opensshd.html

If you'd like, I can test cygwin ssh-agent 6.7p1 with native Windows libssh and see if it actually works. But I do not feel it is appropriate to maintain support for native Windows ssh-agent so long as it is unmaintained.

On Sunday, January 11th 2015 14:27:05, Michael DePaulo wrote:

Back in September, my friend Michael Frederick and I (gmail: psududemike) managed to port that patch to 0.6.3. We've been using it successfully in X2Go Client for Windows 4.0.3.0 and later.
That patch still applies to 0.6.4. I've attached the patch.

Comments are welcome.

I just signed my DCO. Mike Frederick assigned copyright of this patch to me, but if need be, he can sign the DCO also.

However, I am not sure about the DCO from the authors for the original patch. Actually, there were 2 different versions of the original patch:
0001-implement-support-for-putty-s-pageant.patch - which was superseded by:
0001-implement-support-for-putty-s-pageant-0.5.3.patch
And to make things more confusing, there were multiple versions of the 0.5.3 patch.
That ohloh.net link no longer works. But both original patches are available here.
https://projects.kde.org/projects/kdesupport/emerge/repository/revisions/master/show/portage/win32libs/libssh

The 2 authors of those patches appear to be:
Patrick Spendrin <ps//ml`gmx.de>
Patrick von Reth

So at this point, I am not sure how to proceed. Let me ask: Did those 2 authors ever submit the DCO?

and:

On Thursday, January 15th 2015 09:03:58, Andreas Schneider wrote:

Thanks for your work. Copyright doesn't need to be assigned if all people are individuals. The Certificate of Origin is only for people who send patches with Corporate Copyright!

The patch is far from being in a state that I would include it. We need to design a new API for this, similar to pki.c. With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time. So we need a way to build with both and register the backend at the agent engine. We can discuss this on IRC.

On Friday, April 11th 2014 00:07:08, Jon Simons wrote:

I've also written code as described here with the same movitation:

rc = ssh//event//dopoll(event, timeout//msec);

....

const char **lib//err//msg = ssh//get//error(session);

....

if (rc == SSH//ERROR) {
    /** XXX no better way to extract failure reason **/
    #define DISCONN//MSG "Received SSH//MSG//DISCONNECT: 11"
    int graceful//disconnect = (strstr(lib//err//msg, DISCONN//MSG) != NULL);

...

An extended error getter would be useful.

On Thursday, April 10th 2014 10:49:24, Stef Walter wrote:

The work around I'm using is screen-scraping for certain non-translated strings in sshgeterror() message.

On Tuesday, December 24th 2013 02:28:08, Hisashi Kojima wrote:

Thank you for accepting my suggestion. I read and understood the fixed code. I can not immediately upgrade my code base from 0.5.5 to 0.6.0rc3, but will try the new version. Thanks.

On Saturday, December 21st 2013 10:50:40, Andreas Schneider wrote:

This should be fixed in libssh 0.6.0rc3. Could you please try?

On Monday, September 23rd 2013 14:48:15, Dustin Oprea wrote:

Sounds great.

On Monday, September 23rd 2013 13:56:59, Aris Adamantiadis wrote:

We could update the code of the button to issue a google search specific to the libssh website, what do you think ?

On Monday, September 9th 2013 14:58:37, Andreas Schneider wrote:

There is no code behind that button.

If you want to search our mailing list you can use:

http://news.gmane.org/gmane.network.ssh.libssh.general

On Tuesday, July 17th 2012 15:59:50, Andreas Schneider wrote:

Can you please be more precise or add a code example?

and:

• set category to "session"

On Monday, July 15th 2013 21:01:42, Aris Adamantiadis

• changed the subject from "lib_ssh with Dolphin, not the same speed upload/download with sftp" to "Support ftp async upload (was w/ Dolphin, not the same speed upload/download with sftp)"
• unassigned Did G
• set target version to "0.7.0"

On Monday, November 28th 2011 20:19:26, Vineeth Pillai wrote:

patch added

and:

On Tuesday, August 9th 2011 19:53:04, Felix Hupfeld wrote:

when cutting writes into blocks of sshchannelwindow//size(channel) size, the writes work, but read still doesn't.

[2] Authentication successful
[2] Creating a channel 43 with 64000 window and 32768 max packet
[2] Received a CHANNELOPENCONFIRMATION for channel 43:43
[2] Remote window : 32000, maxpacket : 35000
[2] Channel request subsystem success
[2] Read (16384) buffered : 0 bytes. Window: 32000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 32000, writing 32000
[1] channelwrite wrote 32000 bytes
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] Channel receiving 32000 bytes data in 0 (local win=32000 remote win=64000)
[1] placing 32000 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=0 remote win=64000)
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] growing window (channel 43:43) to 128000 bytes
[2] Read (16384) buffered : 15616 bytes. Window: 128000
[2] Adding 128000 bytes to channel (43:43) (from 0 bytes)
E 10308 ssh.cpp:547 Short read: 15616 < 16384
I 10308 ssh.cpp:520 sshchannelwindowsize returns 128000, writing 68000
..\src\comm\ssh//test.cpp(13): error: Value of: conn->Read(buffer, 100000)

Actual: 0

Expected: 100000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 64000, writing 5
[1] channelwrite wrote 5 bytes
[1] channelwrite wrote 34990 bytes
[1] channelwrite wrote 33010 bytes
[2] Channel receiving 5 bytes data in 0 (local win=64000 remote win=60000)
[1] placing 5 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=63995 remote win=60000)
[2] Read (5) buffered : 5 bytes. Window: 63995
[2] growing window (channel 43:43) to 128000 bytes

and:

• unassigned Aris Adamantiadis

On Tuesday, September 20th 2011 08:19:39, Antonio Orefice wrote:

The same with 0.5.2 (1.1MBps)

On Saturday, September 17th 2011 06:44:31, Andreas Schneider wrote:

We will release 0.5.2 today. Please test and report back.

Moving bug to 0.5.3.

and:

• changed target version from "0.5.2" to "0.5.3"

On Friday, September 9th 2011 13:01:35, Antonio Orefice wrote:

0.5.1 gives the same speed as 0.5.0 under kio//sftp, i didn't tested 0.5.2; waiting for archlinux package.

On Thursday, September 8th 2011 17:11:54, Aris Adamantiadis wrote:

I developped a benchmark tool (tests/benchmark) which exposes the speed of sftp/scp/ssh. I do not see significant performance problems on 0.5.2, but maybe it's only affecting 0.5.0.

On Monday, August 29th 2011 09:03:18, Andreas Schneider

• Raised the task's priority to High
• changed target version from "0.5.1" to "0.5.2"

On Friday, July 22nd 2011 12:19:37, Aris Adamantiadis wrote:

Hi Antonio,

Thanks for you feedback. I will check ASAP what's causing this regression. We've made many architectural changes between 0.4 and 0.5 and there is room for such kind of problems.

Kr,

Aris

and:

• set category to "sftp"
• changed task status from "New" to "Assigned"
• assigned the task to Aris Adamantiadis
• Raised the task's priority to Normal
• set target version to "0.5.1"
• set estimated hours to 3

On Tuesday, October 20th 2015 18:21:21, Andreas Schneider

• assigned the task to Aris Adamantiadis

On Tuesday, October 20th 2015 18:21:54, Andreas Schneider

• assigned the task to Andreas Schneider

On Friday, January 1st 2016 21:18:55, quentin garnier wrote:

I get the result of commands with sshchannelread and sshchannelselect. it works well.

I have tried to use events and i have problems:

• channelexitstatus//function seems to not be launched
• channeldatafunction can be launched or not
• i have segmentation fault

My process:

• i connect a session
• open channels: CA and CB
• create event
• add my session in the event
• set callback for channel CA
• execute commands in Channels CA and CB (only the execution with sshchannelrequest//exec)
• open two other channels CC and CD
• call ssheventdopoll to launch callbacks

I have done something wrong ? Maybe you need the trace with verbosity max ? Thanks for the help.

On Thursday, December 31st 2015 13:00:30, quentin garnier wrote:

Ok, i have find a solution for the channel id: i use localchannel integer and remotechannel integer from the struct channel. Well better than the pointer address.

On Monday, December 28th 2015 14:59:26, quentin garnier wrote:

I have a question. I would like to use following functions:

• sshchannelread//timeout
• sshchannelselect

Like that i could launch X channels with X commands in parrallel. But i don't know how to identify the channel.
I don't really want to use pointer value. Not very elegant. If i can have a function to get the stdout fd of a channel.

On Monday, December 28th 2015 12:09:29, quentin garnier wrote:

I have beginning to work on perl binding here:
https://github.com/garnier-quentin/perl-libssh

You're writing in doc:

Warning: sshselect() and sshchannel//select() are not relevant anymore,
since libssh is about to provide an easier system for asynchronous
communications. This subsection should be removed then. **

What we should use instead ?

On Wednesday, December 17th 2014 09:34:53, Anonymous wrote:

I'd like to use ftp instead. But i didn't find a timeout option if nothing is sent (like 'ServerAliveInterval' and 'ServerAliveCountMax' options).

On Wednesday, December 17th 2014 09:29:33, Andreas Schneider wrote:

Why don't you use sftp instead of this horrible scp protocol?

On Friday, January 16th 2015 21:41:54, Michael DePaulo wrote:

Andreas Schneider wrote:

With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time.

I did some research :

1. It appears that there is no actively maintained port of OpenSSH (and therefore ssh-agent) to Windows. In order to not be affected by a vulnerability, one needs to use OpenSSH 6.7p1 or later.

2. In the past, NoMachine ported SSH to Windows. But they are not maintaining it anymore. The last ported version was 5.9p1.

https://www.nomachine.com/AR11K00739
https://www.nomachine.com/NoMachine-OSS-ports

3. There is actively maintained software such as "OpenSSH for Windows", which is at 6.7p1. But they are actually OpenSSH for Cygwin with an easy Windows installer.

http://www.mls-software.com/opensshd.html

If you'd like, I can test cygwin ssh-agent 6.7p1 with native Windows libssh and see if it actually works. But I do not feel it is appropriate to maintain support for native Windows ssh-agent so long as it is unmaintained.

On Thursday, January 15th 2015 09:03:58, Andreas Schneider wrote:

Thanks for your work. Copyright doesn't need to be assigned if all people are individuals. The Certificate of Origin is only for people who send patches with Corporate Copyright!

The patch is far from being in a state that I would include it. We need to design a new API for this, similar to pki.c. With ssh-agens it is possible that you want to support ssh-agend and pagent on Windows at the same time. So we need a way to build with both and register the backend at the agent engine. We can discuss this on IRC.

On Sunday, January 11th 2015 14:27:05, Michael DePaulo wrote:

Back in September, my friend Michael Frederick and I (gmail: psududemike) managed to port that patch to 0.6.3. We've been using it successfully in X2Go Client for Windows 4.0.3.0 and later.
That patch still applies to 0.6.4. I've attached the patch.

Comments are welcome.

I just signed my DCO. Mike Frederick assigned copyright of this patch to me, but if need be, he can sign the DCO also.

However, I am not sure about the DCO from the authors for the original patch. Actually, there were 2 different versions of the original patch:
0001-implement-support-for-putty-s-pageant.patch - which was superseded by:
0001-implement-support-for-putty-s-pageant-0.5.3.patch
And to make things more confusing, there were multiple versions of the 0.5.3 patch.
That ohloh.net link no longer works. But both original patches are available here.
https://projects.kde.org/projects/kdesupport/emerge/repository/revisions/master/show/portage/win32libs/libssh

The 2 authors of those patches appear to be:
Patrick Spendrin <ps//ml`gmx.de>
Patrick von Reth

So at this point, I am not sure how to proceed. Let me ask: Did those 2 authors ever submit the DCO?

and:

On Friday, April 11th 2014 00:07:08, Jon Simons wrote:

I've also written code as described here with the same movitation:

rc = ssh//event//dopoll(event, timeout//msec);

....

const char **lib//err//msg = ssh//get//error(session);

....

if (rc == SSH//ERROR) {
    /** XXX no better way to extract failure reason **/
    #define DISCONN//MSG "Received SSH//MSG//DISCONNECT: 11"
    int graceful//disconnect = (strstr(lib//err//msg, DISCONN//MSG) != NULL);

...

An extended error getter would be useful.

On Thursday, April 10th 2014 10:49:24, Stef Walter wrote:

The work around I'm using is screen-scraping for certain non-translated strings in sshgeterror() message.

On Tuesday, December 24th 2013 02:28:08, Hisashi Kojima wrote:

Thank you for accepting my suggestion. I read and understood the fixed code. I can not immediately upgrade my code base from 0.5.5 to 0.6.0rc3, but will try the new version. Thanks.

On Saturday, December 21st 2013 10:50:40, Andreas Schneider wrote:

This should be fixed in libssh 0.6.0rc3. Could you please try?

On Monday, September 23rd 2013 14:48:15, Dustin Oprea wrote:

Sounds great.

On Monday, September 23rd 2013 13:56:59, Aris Adamantiadis wrote:

We could update the code of the button to issue a google search specific to the libssh website, what do you think ?

On Monday, September 9th 2013 14:58:37, Andreas Schneider wrote:

There is no code behind that button.

If you want to search our mailing list you can use:

http://news.gmane.org/gmane.network.ssh.libssh.general

On Tuesday, July 17th 2012 15:59:50, Andreas Schneider wrote:

Can you please be more precise or add a code example?

and:

• set category to "session"

On Monday, July 15th 2013 21:01:42, Aris Adamantiadis

• changed the subject from "lib_ssh with Dolphin, not the same speed upload/download with sftp" to "Support ftp async upload (was w/ Dolphin, not the same speed upload/download with sftp)"
• unassigned Did G
• set target version to "0.7.0"

On Monday, November 28th 2011 20:19:26, Vineeth Pillai wrote:

patch added

and:

On Tuesday, August 9th 2011 19:53:04, Felix Hupfeld wrote:

when cutting writes into blocks of sshchannelwindow//size(channel) size, the writes work, but read still doesn't.

[2] Authentication successful
[2] Creating a channel 43 with 64000 window and 32768 max packet
[2] Received a CHANNELOPENCONFIRMATION for channel 43:43
[2] Remote window : 32000, maxpacket : 35000
[2] Channel request subsystem success
[2] Read (16384) buffered : 0 bytes. Window: 32000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 32000, writing 32000
[1] channelwrite wrote 32000 bytes
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] Channel receiving 32000 bytes data in 0 (local win=32000 remote win=64000)
[1] placing 32000 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=0 remote win=64000)
I 10308 ssh.cpp:520 sshchannelwindowsize returns 0, writing 0
[2] growing window (channel 43:43) to 128000 bytes
[2] Read (16384) buffered : 15616 bytes. Window: 128000
[2] Adding 128000 bytes to channel (43:43) (from 0 bytes)
E 10308 ssh.cpp:547 Short read: 15616 < 16384
I 10308 ssh.cpp:520 sshchannelwindowsize returns 128000, writing 68000
..\src\comm\ssh//test.cpp(13): error: Value of: conn->Read(buffer, 100000)

Actual: 0

Expected: 100000
I 10308 ssh.cpp:520 sshchannelwindowsize returns 64000, writing 5
[1] channelwrite wrote 5 bytes
[1] channelwrite wrote 34990 bytes
[1] channelwrite wrote 33010 bytes
[2] Channel receiving 5 bytes data in 0 (local win=64000 remote win=60000)
[1] placing 5 bytes into channel buffer (stderr=0)
[2] Channel windows are now (local win=63995 remote win=60000)
[2] Read (5) buffered : 5 bytes. Window: 63995
[2] growing window (channel 43:43) to 128000 bytes

and:

• unassigned Aris Adamantiadis

On Tuesday, September 20th 2011 08:19:39, Antonio Orefice wrote:

The same with 0.5.2 (1.1MBps)

On Saturday, September 17th 2011 06:44:31, Andreas Schneider wrote:

We will release 0.5.2 today. Please test and report back.

Moving bug to 0.5.3.

and:

• changed target version from "0.5.2" to "0.5.3"

On Friday, September 9th 2011 13:01:35, Antonio Orefice wrote:

0.5.1 gives the same speed as 0.5.0 under kio//sftp, i didn't tested 0.5.2; waiting for archlinux package.

On Thursday, September 8th 2011 17:11:54, Aris Adamantiadis wrote:

I developped a benchmark tool (tests/benchmark) which exposes the speed of sftp/scp/ssh. I do not see significant performance problems on 0.5.2, but maybe it's only affecting 0.5.0.

On Monday, August 29th 2011 09:03:18, Andreas Schneider

• Raised the task's priority to High
• changed target version from "0.5.1" to "0.5.2"

On Friday, July 22nd 2011 12:19:37, Aris Adamantiadis wrote:

Hi Antonio,

Thanks for you feedback. I will check ASAP what's causing this regression. We've made many architectural changes between 0.4 and 0.5 and there is room for such kind of problems.

Kr,

Aris

and:

• set category to "sftp"
• changed task status from "New" to "Assigned"
• assigned the task to Aris Adamantiadis
• Raised the task's priority to Normal
• set target version to "0.5.1"
• set estimated hours to 3