Xpra

Xpra New features:
* Clipboard sharing (with full X semantics).
* Icon support.
* Support for raw TCP sockets. Insecure if you don't know what
you are doing.

Xpra Bug fixes:
* Xvfb doesn't support mouse wheels, so they still don't work in
xpra. But now xpra doesn't crash if you try.
* Running FSF Emacs under xpra no longer creates an infinite loop.
* The directory that xpra was launched from is now correctly
saved in ~/.xpra/run-xpra.
* Work around PyGtk weirdness that caused the server and client
to sometimes ignore control-C.
* The client correctly notices keyboard layout changes.
* The client no longer crashes on keymaps in which unnamed keys
are bound to modifiers.
* Workarounds are included for several buggy versions of Pyrex.

Wimpiggy:
* Assume that EWMH-style icons have non-premultiplied alpha.

Other:
* Add copyright comments to all source files.

This release primarily contains cleanups and bugfixes for xpra.

General:
* Logging cleanup -- all logging now goes through the Python
logging framework instead of using raw 'prints'. By default,
debug logging is suppressed, but can be enabled in a fine- or
coarse-grained way.

Xpra:
* Protocol changes; [0.0.5] clients can only be used with v0.0.5
servers, and vice-versa. Use 'xpra upgrade' to upgrade old
servers without losing your session state.
* Man page now included.
Important bug fixes:
* Qt apps formerly could not receive keyboard input due to a focus
handling bug; now fixed.
* Fedora's pygtk2 has mysterious local hacks that broke xpra;
a workaround is now included.
UI improvements:
* 'xpra attach ssh:machine' now works out-of-the-box even if xpra
is not present in the remote machine's PATH, or requires
PYTHONPATH tweaks, or whatever. (The server does still need to
be running on the remote machine, though, of course.)
* Commands that connect to a running xpra server ('attach', 'stop',
etc.) now can generally be used without specifying the name of
the server, assuming only one server is running. (E.g., instead
of 'xpra attach :10', you can use 'xpra attach'; ditto for remote
hosts, you can now use plain 'xpra attach ssh:remote'.)
* Mouse scroll wheels now supported.
* 'xpra start' can now spawn child programs directly (--with-child)
and exit automatically when these children have exited
(--exit-with-children).
Other:
* More robust strategy for handling window stacking order.
(Side-effect: the xpra client no longer requires you to be using
an EWMH-compliant window manager.)
* The xpra client no longer crashes when receiving an unknown key
event (e.g. a multimedia key).
* Very brief transient windows (e.g., tooltips) no longer create
persistent "litter" on the screen.
* Windows with non-empty X borders (e.g., xterm popup menus) are
now handled properly.
* Withdrawn windows no longer reappear after 'xpra upgrade'.

Wimpiggy:
* Do not segfault when querying the tree structure of destroyed
windows.
* Other bugfixes.

Parti:
* No changes.

Xpra:
* Protocol changes break compatibility with 0.0.3, but:
* New command 'xpra upgrade', to restart/upgrade an xpra server
without losing any client state. (Won't work when upgrading from

* Fix bug that left stray busy-looping processes behind on server
when using ssh connections.
* Export window class/instance hints (patch from Ethan Blanton).
* Hack to make backspace key work (full support for keyboard maps
still TBD).
* Added discussion of xmove to README.xpra.

Wimpiggy:
* Make compatible with current Pyrex releases (thanks to many
people for reporting this).
* Work around X server bug 14648 (thanks to Ethan Blanton for help
tracking this down). This improves speed dramatically.
* Reverse-engineer X server lifetime rules for NameWindowPixmap,
and handle it properly. Also handle it lazily. This fixes the
bug where window contents stop updating.
* Avoid crashing when acknowledging damage against an
already-closed window.
* Improve server extension checking (thanks to 'moreilcon' for the
report).
* Remove spurious (and harmless) assertion messages when a window
closes.
* Make manager selection handling fully ICCCM-compliant (in
particular, we now pause properly while waiting for a previous
window manager to exit).
* Make algorithm for classifying unmapped client windows fully
correct.
* Reduce required version of Composite extension to 0.2.

Parti:
* Remove a stale import that caused a crash at runtime (thanks to
'astronouth7303' for the report).

General:
* Error out build with useful error message if required packages
are missing.

Parti 0.0.3 2008-02-20
Massive refactoring occurred for this release.

wimpiggy:
The WM backend parts of Parti have been split off into a
separate package known as wimpiggy. As compared to the corresponding
code in 0.0.2, wimpiggy 0.0.3 adds:
* Compositing support
* Model/view separation for client windows (based on compositing
support)
* Improved client hint support, including icon handling, strut
handling, and more correct geometry handling.
* Keybinding support
* Event dispatching that doesn't leak memory
* Better interaction with already running window managers (i.e., a
--replace switch as seen in metacity etc.)

parti:
This package will eventually become the real window manager,
but for now is essentially a testbed for wimpiggy.

xpra:
This is a new, independent program dependent on wimpiggy (which
is why wimpiggy had to be split out). It implements 'screen for X' --
letting one run applications remotely that can be detached and then
re-attached without losing state. This is the first release, but
while not perfect, it is substantially usable.

general:
The test runner was hacked to share a single X/D-Bus session
across multiple tests. This speeds up the test suite by a factor of
~3, but seems to be buggy and fragile and may be reverted in the
future.


Parti 0.0.2 2007-10-26
This release adds a mostly comprehensive test suite, plus fixes a lot
of bugs. Still only useful for experimentation and hacking.

'python setup.py sdist' sort of works now.


Parti 0.0.1 2007-08-10
Initial release.

Contains basic window manager functionality, including a fair amount
of compliance to ICCCM/EWMH, focus handling, etc., and doesn't seem to
crash in basic testing.

Doesn't do much useful with this; only a simple placeholder layout
manager is included, and only skeleton of virtual desktop support is
yet written.