Clipboard
This feature allows you to copy from outside the xpra session and paste inside it, and vice versa.
For various reasons, this doesn’t always work quite as well as expected - see below for more details.
- Mac OSX and MS Windows clients only have a single clipboard selection whereas X11 has three:
CLIPBOARD, PRIMARY and SECONDARY; therefore we need to choose which one to exchange with: see local-clipboard and remote-clipboard switches
- Mac OSX and MS Windows clients synchronize the remote
PRIMARY selection by default: see selections below
- on MS Windows, the OS requests the clipboard data as soon as we claim ownership
- on MacOS, we have to use polling to see client-side changes
- the HTML5 client can only access the clipboard when the browser decides it is appropriate to do so: usually following clicks or specific key combinations, it may also request permission once and this cannot be easily enabled afterwards once it has been denied
- Wayland severely restricts access to the clipboard, making it impossible to synchronize it properly
Configuration Options
clipboard: auto, yes, no, all, or the name of a clipboard implementationclipboard-direction: can be used to restrict the direction of the clipboard data transfers. This setting is also available from the system tray menu.clipboard-filter-file: can be used to filter out clipboard contents using a file containing regular expressionslocal-clipboard / remote-clipboard: can be used to select which clipboard selection to synchronize with
Selections
The default auto setting synchronizes all the selections supported by the platform.
On MS Windows and MacOS, auto is equivalent to all.
--clipboard=all enables every selection that the platform can handle.
On Mac OSX and MS Windows clients, this adds the PRIMARY selection, which is handled as a special case since it does not exist locally:
- it is only ever synchronized from the server to the client, we never claim it
- the server just notifies us when its contents change, and we then request the data using a slow timer (200ms, configurable with the
XPRA_CLIPBOARD_PRIMARY_DELAY environment variable) so that selecting text with the mouse does not flood the connection
- the data we receive is saved to the local
CLIPBOARD selection, without claiming the remote CLIPBOARD
- a
CLIPBOARD token or a local clipboard change cancels any pending PRIMARY request, so that a real copy always wins over a mouse selection: if the request had already been sent, its reply is discarded instead
This is implemented by PrimaryProxyMixin in xpra/clipboard/primary.py, which both platform backends share.
Technical Constraints
Clipboard support is an ongoing struggle.
You must ensure that there are no other clipboard synchronization tools already running as those are very likely to interfere and cause synchronization loops, wasted bandwidth, trigger application bugs, etc
In particular, avoid using the clipboard synchronization from your virtualization solution if you use one (ie: virtualbox, vmware, etc), or from tools like synergy. Alternatively, you can disable xpra’s clipboard instead. Just avoid running both at the same time.
Debugging
There is a debugging tool which can be launched using xpra clipboard-test.
The easiest way to test the clipboard on X11 platforms is to use xclip tool to set and verify the contents of each clipboard.
- to set the “primary” clipboard contents to the string “primary”:
echo _primary_ | xclip -i -selection primary
- To print the contents of the “primary” clipboard:
xclip -o -selection primary`
Note: on win32, you will need to change the clipboard currently in use to match the one you modify, this must be done before changing the value to ensure it is propagated.
Logging
just add -d clipboard to your xpra command line.
Useful Pointers
For low level implementation details, see clipboard subsystem.
- How does X11 clipboard handle multiple data formats?
- x11-clipboard.cpp from
VirtualBox
- operating system specific clipboards on wikipedia
- The X11 clipboard An overview of its problems and a proposed solution
And here is a good quote from it:
Clipboard sharing and network transparency: It’s nearly impossible to make the clipboard shared across different desktop computers. In fact, it is possible, but such an implementation would be needlessly difficult and complex. The same can be said
of support for virtualization (Qemu, Xen, VMWare). Sharing the clipboard between a virtual machine and the desktop itself is painfully difficult to implement correctly (in case X11 is running on the host operating system).
Related tickets
- #2312 clipboard images with html5 client
- #2634 disable clipboard watermarks
- #1844 async clipboard api
- #41: when we support concurrent users on the same session, we currently give the clipboard to the first client - doing anything else will be quite tricky
- #812 re-implement clipboard without gtk or nested main
- #1167 tray menu clipboard choice irreversible
- #1139 XPRA - Matlab - Clipboard blinking, UI unresponsive
- #966 provide a persistent setting to select the default clipboard to synchronize
- #1112 clipboard notification flashing constantly
- #1018 recursion depth error
- #883 Pasting into WYSIWYG editors rich text fails and causes too many clipboard requests
- #842 html5 client clipboard support
- #834 Sync issue with win32 client clipboard
- #823 Session hangs and dies
- #877 clipboard hitting maximum requests per second limit
- #735 Clipboard working incorrect with java-applications (x2go + xpra)
- #703 Copying URL from web browser address bar on remote host fails
- #272 win32 multiple clipboards enhancement
- #318 osx client-to-server clipboard support
- #273 handle more clipboard formats
- #274 advanced clipboard filtering
- #275 handle clipboard large data transfers better
- #276 limit clipboard direction
- #452 detect and avoid creating clipboard loops
- #313 speedup paste
- #184 clipboard related bug, clipboard can fire at any time… so bugs may appear to come from somewhere else when in fact it is the clipboard that is the source of the problem - to keep in mind
- #162 very hard to reproduce bug - relied on the list (and their order) of X11 atoms defined, as we tried to parse invalid values as X11 atoms
- #176 32-bit vs 64-bit structures issue
- #156 we drop clipboard packets that are too big - rather than causing network problems (bug was that we dropped the connection instead - oops!)
- #52 another atom name issue, this time with Java apps
- #8, #84 and #99 (dupe: #104): more clipboard atom problems
- #11 win32 and osx clipboard ticket (old)