Recent changes to bugs

rfbScreenCleanup() free frameBuffer issue.

Anonymous — Wed, 28 Nov 2012 09:12:26 -0000

in rfbScreenCleanup(), free(ptr->frameBuffer); while not check null pointer. This may cause crash.
Because in my Android server example, to improve performance, I directly set screen->frameBuffer as fbmmap!
So in cleanup(), I munmap(fbmmap), then this memory pointer is not valid anymore. And then crash occurs while followed invoking rfbScreenCleanup().
To avoid crash, after munmap(fbmmap), I have to allocate some bytes memory and assign it to screen->frameBuffer for free.
I think in rfbScreenCleanup(), or do not free frameBuffer and let it free by userself or check it's validity before call free.

Unknown authentication scheme

jobi — Tue, 02 Oct 2012 10:28:59 -0000

From Ubuntu 12.04 trying to connect with remmina to win servers with free RealVNC 4 and we have the errors -> 1) Unknown authentication scheme from VNC server: 5
2) Another error is : Too many security failures

This would be very helpful for us if it was fixed.
Thank you.

Unknown authentication scheme

jobi — Tue, 02 Oct 2012 06:53:42 -0000

This would be very helpful for us if it was fixed.
Thank you.

noVNC not working with iOS Safari and desktop Safari v5

Anonymous — Tue, 07 Aug 2012 21:09:59 -0000

none of the simple example works with noVNC if using safari 5 (e.g. ipad 3 browser). Safari 6 (shipped with Lion) works fine.

Chrome/Firefox works ok.

noVNC connects ok, but there is no screen shown up after connection.

socket not work on windows

Anonymous — Wed, 18 Jul 2012 05:59:08 -0000

I built libvnc on ubuntu with mingw32, and I run example/example.exe on both Windows XP and Windows 7. I got the following error:
httpCheckFds: select: Unknown error
rfbCheckFds: select: Unknow error,
WSAGetLastError() return 10038.

the reason is for Windows, SOCKET was not defined int, it is unsigned int. you can find it in
/usr/i586-mingw32msvc/include/winsock2.h, (on my ubuntu) or
ftp://ftp.microsoft.com/bussys/wINSOCK/winsock2/winsock2.h

typedef unsigned int u_int;
typedef unsigned long u_long;

/*
* The new type to be used in all
* instances which refer to sockets.
*/
#ifdef _WIN64
typedef UINT_PTR SOCKET;
#else
typedef u_int SOCKET;
#endif

So, the following code will always get to run：
if (rfbScreen->httpSock >= 0) {
FD_SET(rfbScreen->httpSock, &fds);
}
even the value of httpSock is -1.
--- a/rfb/rfb.h
+++ b/rfb/rfb.h
I did the following chang:
#ifdef __MINGW32__
#undef SOCKET
#include <winsock2.h>
+#undef SOCKET
+#define SOCKET int
#ifdef LIBVNCSERVER_HAVE_WS2TCPIP_H
#undef socklen_t
#include <ws2tcpip.h>

and tested on both windows7 and windows xp. it works.
but if some new file don't include rfb.h, then it will met this bug again.
in winsock2.h
#define INVALID_SOCKET (SOCKET)(~0),
it is -1.
So we'd better to change all of the if condition test, to compare with INVALID_SOCKET. not >= 0.

and another issue is, on Windows, we can't use read to recv packets. we have to use recv. So I also got
httpProcessInput: read: No error
I did following changes:
--- a/libvncserver/httpd.c
+++ b/libvncserver/httpd.c
@@ -300,8 +300,8 @@ httpProcessInput(rfbScreenInfoPtr rfbScreen)
return;
}

- got = read (rfbScreen->httpSock, buf + buf_filled,
- sizeof (buf) - buf_filled - 1);
+ got = recv (rfbScreen->httpSock, buf + buf_filled,
+ sizeof (buf) - buf_filled - 1, 0);

I tested on both Windows XP, and Windows7, it works.

all of my testing are only run example. Thanks.

I also upload my diff.

libvncserver not accepting link local addresses

Oliver — Fri, 13 Jul 2012 17:36:53 -0000

Hello,

I'm just adding IPv6 support to VirtualBox and when trying to bind to a link local IPv6 address (the fe80:: ones), I can see this error:

[X] orgasmatron: today root$ VBoxHeadless -s vmvboxdevel --vrde config
Oracle VM VirtualBox Headless Interface 4.1.51_OSE
(C) 2008-2012 Oracle Corporation
All rights reserved.

IPv4: Address: (null), Port: 5999
IPv6: Address: fe80::5604:a6ff:fef1:73c6, Port: 5903
13/07/2012 19:31:56 Listening for VNC connections on TCP port 5999
13/07/2012 19:31:56 rfbListenOnTCP6Port: error in bind IPv6 socket: Invalid argument
VRDE server is listening on port 5999.
^C
[X] orgasmatron: today root$

When changing the address to a global one:

[X] orgasmatron: today root$ VBoxHeadless -s vmvboxdevel --vrde config
Oracle VM VirtualBox Headless Interface 4.1.51_OSE
(C) 2008-2012 Oracle Corporation
All rights reserved.

IPv4: Address: (null), Port: 5999
IPv6: Address: 2001:db8::dead:beef, Port: 5903
13/07/2012 19:36:05 Listening for VNC connections on TCP port 5999
13/07/2012 19:36:05 Listening for VNC connections on TCP6 port 5903
VRDE server is listening on port 5999.
^C
[X] orgasmatron: today root$

It's working. Bug or Feature?

KR,

Oliver

SONAME should be bumped

Luca Falavigna — Thu, 06 Oct 2011 22:04:17 -0000

Recent changes in libvncserver (i.e. commit 0a4f1bad, rfb/rfbclient.h) break ABI and could lead to some segfaults:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffe8ff4700 (LWP 4509)]
0x00007fffee40fc70 in ?? () from /usr/lib/libvncclient.so.0
(gdb) bt
#0 0x00007fffee40fc70 in ?? () from /usr/lib/libvncclient.so.0
#1 0x00007fffee418d5c in HandleRFBServerMessage () from /usr/lib/libvncclient.so.0
#2 0x00007fffee629a9b in ?? () from /usr/lib/remmina/plugins/remmina-plugin-vnc.so
#3 0x00007fffee62c142 in ?? () from /usr/lib/remmina/plugins/remmina-plugin-vnc.so
#4 0x00007fffee62c1e6 in ?? () from /usr/lib/remmina/plugins/remmina-plugin-vnc.so
#5 0x00007ffff43a6b40 in start_thread () from /lib/x86_64-linux-gnu/libpthread.so.0
#6 0x00007ffff40f136d in clone () from /lib/x86_64-linux-gnu/libc.so.6
#7 0x0000000000000000 in ?? ()

You should consider bumping SONAME to address this.

Also see http://anonscm.debian.org/gitweb/?p=users/dktrkranz/abi.git;a=tree;f=libvncserver/0.9.7_0.9.8 for further reference

libvncserver 0.9.7 password length bug

Anonymous — Wed, 22 Dec 2010 20:34:58 -0000

In libvncserver-0.9.7, when using rfbScreen->passwordCheck = rfbCheckPasswordByList, a password that is longer than the real password can be entered and as long as it starts with the original password it is accepted. This is also true when using a custom passwordCheck and rfbEncryptBytes.

Example:

Answer = password
Guess = password
ACCEPTED

Answer = password
Guess = passwor
DENIED

Answer = password
Guess = passwordsjg923trin8yrcgnubxyqweqe9qw87e9r8
ACCEPTED

Brandon Holland
http://brandon-holland.com

32 bpp and 32 depth SEGFAULT

Anonymous — Mon, 29 Nov 2010 14:36:36 -0000

29/11/2010 12:28:05 Pixel format for client 192.168.1.138:
29/11/2010 12:28:05 32 bpp, depth 32, little endian
29/11/2010 12:28:05 true colour: max r 255 g 255 b 255, shift r 0 g 8 b 16
29/11/2010 12:28:05 no translation needed
29/11/2010 12:28:05 rfbProcessClientNormalMessage: ignoring unsupported encoding type zlibhex
29/11/2010 12:28:05 Using ZRLE encoding for client 192.168.1.138
Segmentation fault

When i a try to start a vncServer with 32bpp and the client connects i get a SEGFAULT from the library

if (0)

yerp — Sun, 15 Aug 2010 18:33:50 -0000

Continuation of: https://sourceforge.net/tracker/?func=detail&atid=405858&aid=3044889&group_id=32584
It seems you closed that post and I cannot reply or submit a patch.
The problem is not solved.

I only found out about this if (0) case because of a linker error from code included inside the if (0).
This does not get optimized away as you claim.
I am using GCC 4.4.4 and Binutils 2.20.

If you do not close this, I can make a patch that changes all of those if (0) cases into comments when I get around to it.