- As with the encoder, the decoder now uses the TurboJPEG wrapper, which
allows it to decode JPEG images directly into the framebuffer. This
eliminates a buffer copy (CopyRectangle()) as well as the expensive
RGB pixel conversion in DecompressJpegRectBPP(). The TurboJPEG
wrapper performs RGB pixel conversion more optimally, and only when
necessary (it uses the libjpeg-turbo colorspace extensions when
available, in order to avoid RGB conversion.)
- The other Tight subencoding types are also now decoded directly into
the framebuffer, which eliminates buffer copies.
- The Tight decoder now supports the rfbTightNoZlib extension, which
allows the server to bypass zlib compression when Compression Level 0
is selected. The encoder already supports this extension. Passing
the data stream through zlib when Compression Level 0 is selected
needlessly wastes CPU time, since all zlib is doing is copying the
data internally into its own structures.
It turns out some server implementations (namely VMware ESXi 6.5) expect
1 as the only non-zero value for the SetPixelFormat message whereas the protocol
states every non-zero value is valid
(https://github.com/rfbproto/rfbproto/blob/master/rfbproto.rst#setpixelformat).
Anyway, setting this to 1 shouldn't hurt.
Fixes #141
To support Microsoft Visual C++, we must not guard Windows-specific code
in MinGW-specific #ifdef guards.
Happily, even 64-bit MSVC defines the WIN32 constant, therefore we can use
that instead.
[JES: fixed commit message, reordered commit, split out unrelated changes]
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
If MallocFrameBuffer() returns FALSE, frame buffer pointer is left to
NULL. Subsequent writes into that buffer could lead to memory
corruption, or even arbitrary code execution.
rfbClientSetClientData() allocates a new rfbClientData, but never gets
cleaned up, which causes memory leaks.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This patch implements support in LibVNCClient for framebuffer updates
encoded as H.264 frames. Hardware accelerated decoding is performed
using VA API.
This is experimental support to let the community explore the possibilities
offered by the potential bandwidth and latency reductions that H.264 encoding
allows. This may be particularly useful for use cases such as online gaming,
hosted desktops, hosted set top boxes...
This patch only provides the client side support and is meant to be used
with corresponding server-side support, as provided by an upcoming patch for
qemu ui/vnc module (to view the display of a virtual machine executing under
QEMU).
With this H.264-based encoding, if multiple framebuffer update messages
are generated for a single server framebuffer modification, the H.264
frame data is sent only with the first update message. Subsequent update
framebuffer messages will contain only the coordinates and size of the
additional updated regions.
Instructions/Requirements:
* The patch should be applied on top of the previous patch I submitted with
minor enhancements to the gtkvncviewer application:
http://sourceforge.net/mailarchive/message.php?msg_id=30323804
* Currently only works with libva 1.0: use branch "v1.0-branch" for libva and
intel-driver. Those can be built as follows:
cd libva
git checkout v1.0-branch
./autogen.sh
make
sudo make install
cd ..
git clone git://anongit.freedesktop.org/vaapi/intel-driver
cd intel-driver
git checkout v1.0-branch
./autogen.sh
make
sudo make install
Signed-off-by: David Verbeiren <david.verbeiren@intel.com>
Besided making libvncserver reverseVNC IPv6-aware, this introduces some changes
on the client side as well to make clients listen on IPv6 sockets, too. Like
the server side, this also uses a separate-socket approach.
For security reasons, it can be important to limit which IP addresses a
LibVNCClient-based client should listen for reverse connections. This
commit adds that option.
To preserve binary backwards-compatibility, the field was added to the end
of the rfbclient struct, and the function ListenAtTcpPort retains its
signature (but calls the new ListenAtTcpPortAndAddress).
[jes: shortened the commit subject, added a longer explanation in the
commit body and adjusted style]
Signed-off-by: Luca Stauble <gnekoz@gmail.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This enables setting the DSCP/Traffic Class field of IP/IPv6 packets
sent by a client. For example starting a client with -qosdscp 184
marks all outgoing traffic for expedited forwarding.
Implementation for Win32 is still a TODO, though. See
http://betelco.blogspot.com/2009/03/dscp-marking-under-windows-at.html
for an overview of the Win32 QoS API mess...
The hook is still called after InitialiseRFBConnection() so we can choose
the color settings depending on the vnc server (or settings) in that hook.
This way one can use the "VNC server default format" pixelformat if the
client supports it, or perform a workaround (Intel AMT KVM "classic vnc"
server only works using 8bit colors in RFB3.8)
Signed-off-by: Wouter Van Meir <wouter.vanmeir@pandora.be>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The serverPort member is already set in rfbGetClient(), if we
set it again in rfbInitClient(), this breaks playing of vncrec
files (this relies on serverPort set to -1).
Signed-off-by: Christian Beier <dontmind@freeshell.org>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
[jes: adjusted coding style, made sure port is initialized correctly]
Signed-off-by: Vic Lee <llyzs@163.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
To make this work, we also have to initialize sockets
to a default value of -1.
Also close a client listen socket if it's open.
Signed-off-by: Christian Beier <dontmind@freeshell.org>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
fixes warning about closesocket being implicitly declared.
Signed-off-by: Christian Beier <dontmind@freeshell.org>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Forking the whole process from deep within a library call does
not really work at all with apps that use multiple threads, i.e. every
reasonably modern GUI app. So, provide a non-forking listen function so
that the caller can decide if to fork, start a thread, etc.
This implementation adds a timeout parameter to be able to call the
listen function multiple times so that it's possible to do sth. else
in between, e.g. abort listening.
Signed-off-by: Christian Beier <dontmind@freeshell.org>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When working on a program which searches the display for some image, one
does not want to search again without getting an FB update. Add a
callback to make this possible.
There was an unnecessary assignment, and an assignment of a string that
was to be free()ed later, so it has to be strdup()ed.
Both issues spotted by Roman Held.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Slávek Banko
Tobias Junghans
Michele Calgaro
DRC
simon
dborth
Christian Beier
Balazs Ludmany
gbdj
Timothy Pearson
Daniel Cohen Gindi
Johannes Schindelin
newsoft
Robbert Klarenbeek
David Verbeiren
Luca Stauble
Wouter Van Meir
Vic Lee
Alexander Dorokhine
Stefan Becker
runge
dscho
steven_carr