mutter/README

417 lines
16 KiB
Plaintext
Raw Permalink Normal View History

The original codebase named "Metacity" is not a meta-City as in an
urban center, but rather Meta-ness as in the state of being
meta. i.e. metacity : meta as opacity : opaque. Also it may have
something to do with the Meta key on UNIX keyboards.
2001-07-25 01:43:41 -04:00
Since then, it has been renamed mutter after a rebase on top of
clutter as a compositing manager.
COMPILING MUTTER
2001-07-28 02:11:15 -04:00
===
You need GTK+ 2.2. For startup notification to work you need
libstartup-notification at
http://www.freedesktop.org/software/startup-notification/ or on the
GNOME ftp site.
You need Clutter 1.0. You need gobject-introspection 0.6.3.
2001-07-28 02:11:15 -04:00
REPORTING BUGS AND SUBMITTING PATCHES
===
Report new bugs on http://bugzilla.gnome.org. Please check for
duplicates, *especially* if you are reporting a feature request.
Please do *not* add "me too!" or "yes I really want this!" comments to
feature requests in bugzilla. Please read
http://pobox.com/~hp/features.html prior to adding any kind of flame
about missing features or misfeatures.
2001-07-28 02:11:15 -04:00
Feel free to send patches too; Metacity is relatively small and
simple, so if you find a bug or want to add a feature it should be
pretty easy. Send me mail, or put the patch in bugzilla.
2001-07-28 02:11:15 -04:00
See the HACKING file for some notes on hacking Mutter.
2001-07-28 02:35:19 -04:00
MUTTER FEATURES
===
2001-07-25 01:43:41 -04:00
- Uses GTK+ 2.0 for drawing window frames. This means colors, fonts,
etc. come from GTK+ theme.
- Does not expose the concept of "window manager" to the user. Some
of the features in the GNOME control panel and other parts of the
desktop happen to be implemented in metacity, such as changing your
window border theme, or changing your window navigation shortcuts,
but the user doesn't need to know this.
- Includes only the window manager; does not try to be a desktop
environment. The pager, configuration, etc. are all separate and
modular. The "libwnck" library (which I also wrote) is available
for writing metacity extensions, pagers, and so on. (But libwnck
isn't metacity specific, or GNOME-dependent; it requires only GTK,
and should work with KWin, fvwm2, and other EWMH-compliant WMs.)
- Has a simple theme system and a couple of extra themes come with it.
Change themes via gsettings:
gsettings set org.gnome.desktop.wm.preferences theme Crux
gsettings set org.gnome.desktop.wm.preferences theme Gorilla
gsettings set org.gnome.desktop.wm.preferences theme Atlanta
gsettings set org.gnome.desktop.wm.preferences theme Bright
See theme-format.txt for docs on the theme format. Use
metacity-theme-viewer to preview themes.
- Change number of workspaces via gsettings:
gsettings set org.gnome.desktop.wm.preferences num-workspaces 5
Can also change workspaces from GNOME 2 pager.
- Change focus mode:
gsettings set org.gnome.desktop.wm.preferences focus-mode mouse
gsettings set org.gnome.desktop.wm.preferences focus-mode sloppy
gsettings set org.gnome.desktop.wm.preferences focus-mode click
2001-07-25 01:43:41 -04:00
- Global keybinding defaults include:
Alt-Tab forward cycle window focus
Alt-Shift-Tab backward cycle focus
Alt-Ctrl-Tab forward cycle focus among panels
Alt-Ctrl-Shift-Tab backward cycle focus among panels
Alt-Escape cycle window focus without a popup thingy
Ctrl-Alt-Left Arrow previous workspace
Ctrl-Alt-Right Arrow next workspace
Ctrl-Alt-D minimize/unminimize all, to show desktop
2001-07-25 01:43:41 -04:00
Change keybindings for example:
gsettings set org.gnome.desktop.wm.keybindings switch-to-workspace-1 '[<Alt>F1]'
Also try the GNOME keyboard shortcuts control panel.
2001-07-25 01:43:41 -04:00
- Window keybindings:
2001-07-25 01:43:41 -04:00
Alt-space window menu
Mnemonics work in the menu. That is, Alt-space then underlined
letter in the menu item works.
Choose Move from menu, and arrow keys to move the window.
2001-07-25 23:14:45 -04:00
2001-07-25 01:43:41 -04:00
While moving, hold down Control to move slower, and
Shift to snap to edges.
2001-07-25 23:14:45 -04:00
2001-07-25 01:43:41 -04:00
Choose Resize from menu, and nothing happens yet, but
eventually I might implement something.
2001-07-25 01:43:41 -04:00
Keybindings for things like maximize window, vertical maximize,
etc. can be bound, but may not all exist by default. See
metacity.schemas.
2001-07-25 01:43:41 -04:00
- Window mouse bindings:
Clicking anywhere on frame with button 1 will raise/focus window
If you click a window control, such as the close button, then the
control will activate on button release if you are still over it
on release (as with most GUI toolkits)
If you click and drag borders with button 1 it resizes the window
If you click and drag the titlebar with button 1 it moves the
window.
If you click anywhere on the frame with button 2 it lowers the
window.
2001-07-25 01:43:41 -04:00
If you click anywhere on the frame with button 3 it shows the
window menu.
If you hold down Super (windows key) and click inside a window, it
will move the window (buttons 1 and 2) or show menu (button 3).
Or you can configure a different modifier for this.
2001-07-25 01:43:41 -04:00
2001-07-28 01:37:57 -04:00
If you pick up a window with button 1 and then switch workspaces
the window will come with you to the new workspace, this is
2001-07-25 01:43:41 -04:00
a feature copied from Enlightenment.
2001-07-26 01:10:01 -04:00
If you hold down Shift while moving a window, the window snaps
to edges of other windows and the screen.
2001-07-25 01:43:41 -04:00
- Session management:
Mutter connects to the session manager and will set itself up to
2001-07-28 02:11:15 -04:00
be respawned. It theoretically restores sizes/positions/workspace
for session-aware applications.
2001-07-25 01:43:41 -04:00
- Mutter implements much of the EWMH window manager specification
from freedesktop.org, as well as the older ICCCM. Please refer to
the COMPLIANCE file for information on mutter compliance with
these standards.
2001-07-25 01:43:41 -04:00
2001-07-25 23:14:45 -04:00
- Uses Pango to render text, so has cool i18n capabilities.
Supports UTF-8 window titles and such.
- There are simple animations for actions such as minimization,
to help users see what is happening. Should probably
have a few more of these and make them nicer.
- if you have the proper X setup, set the GDK_USE_XFT=1
environment variable to get antialiased window titles.
- considers the panel when placing windows and maximizing
them.
- handles the window manager selection from the ICCCM. Will exit if
another WM claims it, and can claim it from another WM if you pass
the --replace argument. So if you're running another
ICCCM-compliant WM, you can run "mutter --replace" to replace it
with Metacity.
- does basic colormap handling
- and much more! well, maybe not a lot more.
2003-01-10 00:02:20 -05:00
HOW TO ADD EXTERNAL FEATURES
===
You can write a mutter "plugin" such as a pager, window list, icon
2003-01-10 00:02:20 -05:00
box, task menu, or even things like "window matching" using the
Extended Window Manager Hints. See http://www.freedesktop.org for the
EWMH specification. An easy-to-use library called "libwnck" is
available that uses the EWMH and is specifically designed for writing
WM accessories.
You might be interested in existing accessories such as "Devil's Pie"
by Ross Burton, which add features to Mutter (or other
2003-01-10 00:02:20 -05:00
EWMH-compliant WMs).
MUTTER BUGS, NON-FEATURES, AND CAVEATS
2001-07-25 01:43:41 -04:00
===
See bugzilla: http://bugzilla.gnome.org/query.cgi
2001-07-25 23:14:45 -04:00
2001-07-25 01:43:41 -04:00
FAQ
===
Q: Will you add my feature?
A: If it makes sense to turn on unconditionally, or is genuinely a
harmless preference that I would not be embarrassed to put in a
simple, uncluttered, user-friendly configuration dialog.
2001-07-25 23:14:45 -04:00
If the only rationale for your feature is that other window
managers have it, or that you are personally used to it, or
something like that, then I will not be impressed. Metacity is
firmly in the "choose good defaults" camp rather than the "offer 6
equally broken ways to do it, and let the user pick one" camp.
2001-07-25 01:43:41 -04:00
2001-07-25 23:14:45 -04:00
This is part of a "no crackrock" policy, despite some exceptions
2001-07-25 01:43:41 -04:00
I'm mildly embarrassed about. For example, multiple workspaces
probably constitute crackrock, they confuse most users and really
are not that useful if you have a decent tasklist and so on. But I
am too used to them to turn them off. Or alternatively
iconification/tasklist is crack, and workspaces/pager are good. But
having both is certainly a bit wrong. Sloppy focus is probably
crackrock too.
2001-07-25 01:43:41 -04:00
But don't think unlimited crack is OK just because I slipped up a
little. No slippery slope here.
2001-07-25 23:14:45 -04:00
Don't let this discourage patches and fixes - I love those. ;-)
Just be prepared to hear the above objections if your patch adds
some crack-ridden configuration option.
2001-07-25 01:43:41 -04:00
http://pobox.com/~hp/free-software-ui.html
http://pobox.com/~hp/features.html
Q: Will Mutter be part of GNOME?
2001-07-30 23:59:55 -04:00
A: It is not officially part of GNOME as of GNOME 2.27. We are
hoping to have mutter officially included as of GNOME 2.28.
Q: Why does Mutter remember the workspace/position of some apps
but not others across logout/login?
2001-07-28 01:37:57 -04:00
A: Mutter only stores sizes/positions for apps that are session
managed. As far as I can determine, there is no way to attempt to
remember workspace/position for non-session-aware apps without
2001-07-28 01:37:57 -04:00
causing a lot of weird effects.
The reason is that you don't know which non-SM-aware apps were
launched by the session. When you initially log in, Metacity sees a
bunch of new windows appear. But it can't distinguish between
windows that were stored in your session, or windows you just
launched after logging in. If Metacity tried to guess that a window
2001-07-28 02:11:15 -04:00
was from the session, it could e.g. end up maximizing a dialog, or
put a window you just launched on another desktop or in a weird
2001-07-28 01:37:57 -04:00
place. And in fact I see a lot of bugs like this in window managers
that try to handle non-session-aware apps.
However, for session-aware apps, Mutter can tell that the
2001-07-28 01:37:57 -04:00
application instance is from the session and thus restore it
reliably, assuming the app properly restores the windows it had
open on session save.
So the correct way to fix the situation is to make apps
session-aware. libSM has come with X for years, it's very
standardized, it's shared by GNOME and KDE - even twm is
session-aware. So anyone who won't take a patch to add SM is more
archaic than twm - and you should flame them. ;-)
Docs on session management:
http://www.fifi.org/doc/xspecs/xsmp.txt.gz
http://www.fifi.org/doc/xspecs/SMlib.txt.gz
2001-07-28 01:37:57 -04:00
2001-07-28 02:11:15 -04:00
See also the ICCCM section on SM. For GNOME apps, use the
GnomeClient object. For a simple example of using libSM directly,
twm/session.c in the twm source code is pretty easy to understand.
2001-07-28 01:37:57 -04:00
Q: How about adding viewports in addition to workspaces?
A: I could conceivably be convinced to use viewports _instead_ of
workspaces, though currently I'm not thinking that. But I don't
think it makes any sense to have both; it's just confusing. They
are functionally equivalent.
You may think this means that you won't have certain keybindings,
or something like that. This is a misconception. The only
_fundamental_ difference between viewports and workspaces is that
with viewports, windows can "overlap" and appear partially on
one and partially on another. All other differences that
traditionally exist in other window managers are accidental -
the features commonly associated with viewports can be implemented
for workspaces, and vice versa.
So I don't want to have two kinds of
workspace/desktop/viewport/whatever, but I'm willing to add
features traditionally associated with either kind if those
features make sense.
Q: Why is the panel always on top?
A: Because it's a better user interface, and until we made this not
configurable a bunch of apps were not getting fixed (the app
authors were just saying "put your panel on the bottom" instead of
properly supporting fullscreen mode, and such).
rationales.txt has the bugzilla URL for some flamefesting on this,
if you want to go back and relive the glory.
Read these and the bugzilla stuff before asking/commenting:
http://pobox.com/~hp/free-software-ui.html
http://pobox.com/~hp/features.html
Q: Why is there no edge flipping?
A: This one is also in rationales.txt. Because "ouija board" UI, where
you just move the mouse around and the computer guesses what you
mean, has a lot of issues. This includes mouse focus, shade-hover
mode, edge flipping, autoraise, etc. Metacity has mouse focus and
autoraise as a compromise, but these features are all confusing for
many users, and cause problems with accessibility, fitt's law, and
so on.
Read these and the bugzilla stuff before asking/commenting:
http://pobox.com/~hp/free-software-ui.html
http://pobox.com/~hp/features.html
Merge reduced_resources mode patch from the branch. Offers wireframe and 2003-10-12 Havoc Pennington <hp@redhat.com> Merge reduced_resources mode patch from the branch. Offers wireframe and no-animations. * src/window.c (implement_showing): no animation if we are in reduced resources mode * src/prefs.c: add REDUCED_RESOURCES pref * src/window.c (meta_window_update_keyboard_resize): fix to modify grab_anchor_window_pos to grab_wireframe_rect if appropriate instead of window->rect * src/display.h (struct _MetaDisplay): add grab_start_serial used to avoid responding to events that occurred prior to the grab initialization. Still broken in various ways, specifically EnterNotify that occurred prior to XGrabPointer is processed as if it occurred after. * src/window.c (meta_window_update_keyboard_move): add this instead of meta_window_warp_pointer() crack * src/effects.c (meta_effects_update_wireframe): draw a kind of grid for the wireframe, instead of just a rectangle, like twm * src/screen.c (meta_screen_new): line width of 3 for the XOR gc "Reduced resources" mode based on wireframe patch from Erwann Chenede. Still pretty buggy. * src/keybindings.c (process_keyboard_move_grab) (process_keyboard_resize_grab): add gruesome wireframe hacks * src/display.c (meta_display_end_grab_op): end wireframe (meta_display_begin_grab_op): begin wireframe * src/effects.c (meta_effects_end_wireframe) (meta_effects_update_wireframe, meta_effects_begin_wireframe): routines to draw the wireframe stuff * src/window.c (window_should_be_showing): hide window when doing wireframe, commented out as it breaks grab * src/window.c (meta_window_refresh_resize_popup): handle wireframe * src/screen.c (meta_screen_new): create a screen->root_xor_gc for use in drawing wireframes * src/frames.c (meta_frames_push_delay_exposes): repaint everything before we delay
2003-10-12 02:25:38 -04:00
Q: Why does wireframe move/resize suck?
Merge reduced_resources mode patch from the branch. Offers wireframe and 2003-10-12 Havoc Pennington <hp@redhat.com> Merge reduced_resources mode patch from the branch. Offers wireframe and no-animations. * src/window.c (implement_showing): no animation if we are in reduced resources mode * src/prefs.c: add REDUCED_RESOURCES pref * src/window.c (meta_window_update_keyboard_resize): fix to modify grab_anchor_window_pos to grab_wireframe_rect if appropriate instead of window->rect * src/display.h (struct _MetaDisplay): add grab_start_serial used to avoid responding to events that occurred prior to the grab initialization. Still broken in various ways, specifically EnterNotify that occurred prior to XGrabPointer is processed as if it occurred after. * src/window.c (meta_window_update_keyboard_move): add this instead of meta_window_warp_pointer() crack * src/effects.c (meta_effects_update_wireframe): draw a kind of grid for the wireframe, instead of just a rectangle, like twm * src/screen.c (meta_screen_new): line width of 3 for the XOR gc "Reduced resources" mode based on wireframe patch from Erwann Chenede. Still pretty buggy. * src/keybindings.c (process_keyboard_move_grab) (process_keyboard_resize_grab): add gruesome wireframe hacks * src/display.c (meta_display_end_grab_op): end wireframe (meta_display_begin_grab_op): begin wireframe * src/effects.c (meta_effects_end_wireframe) (meta_effects_update_wireframe, meta_effects_begin_wireframe): routines to draw the wireframe stuff * src/window.c (window_should_be_showing): hide window when doing wireframe, commented out as it breaks grab * src/window.c (meta_window_refresh_resize_popup): handle wireframe * src/screen.c (meta_screen_new): create a screen->root_xor_gc for use in drawing wireframes * src/frames.c (meta_frames_push_delay_exposes): repaint everything before we delay
2003-10-12 02:25:38 -04:00
A: You can turn it on with the reduced_resources setting.
Merge reduced_resources mode patch from the branch. Offers wireframe and 2003-10-12 Havoc Pennington <hp@redhat.com> Merge reduced_resources mode patch from the branch. Offers wireframe and no-animations. * src/window.c (implement_showing): no animation if we are in reduced resources mode * src/prefs.c: add REDUCED_RESOURCES pref * src/window.c (meta_window_update_keyboard_resize): fix to modify grab_anchor_window_pos to grab_wireframe_rect if appropriate instead of window->rect * src/display.h (struct _MetaDisplay): add grab_start_serial used to avoid responding to events that occurred prior to the grab initialization. Still broken in various ways, specifically EnterNotify that occurred prior to XGrabPointer is processed as if it occurred after. * src/window.c (meta_window_update_keyboard_move): add this instead of meta_window_warp_pointer() crack * src/effects.c (meta_effects_update_wireframe): draw a kind of grid for the wireframe, instead of just a rectangle, like twm * src/screen.c (meta_screen_new): line width of 3 for the XOR gc "Reduced resources" mode based on wireframe patch from Erwann Chenede. Still pretty buggy. * src/keybindings.c (process_keyboard_move_grab) (process_keyboard_resize_grab): add gruesome wireframe hacks * src/display.c (meta_display_end_grab_op): end wireframe (meta_display_begin_grab_op): begin wireframe * src/effects.c (meta_effects_end_wireframe) (meta_effects_update_wireframe, meta_effects_begin_wireframe): routines to draw the wireframe stuff * src/window.c (window_should_be_showing): hide window when doing wireframe, commented out as it breaks grab * src/window.c (meta_window_refresh_resize_popup): handle wireframe * src/screen.c (meta_screen_new): create a screen->root_xor_gc for use in drawing wireframes * src/frames.c (meta_frames_push_delay_exposes): repaint everything before we delay
2003-10-12 02:25:38 -04:00
But: it has low usability, and is a pain
to implement, and there's no reason opaque move/resize should be a
problem on any setup that can run a modern desktop worth a darn to
begin with.
Read these and the bugzilla stuff before asking/commenting:
http://pobox.com/~hp/free-software-ui.html
http://pobox.com/~hp/features.html
The reason we had to add wireframe anyway was broken
proprietary apps that can't handle lots of resize events.
Q: Why no XYZ?
A: You are probably getting the idea by now - check rationales.txt,
query/search bugzilla, and read http://pobox.com/~hp/features.html
and http://pobox.com/~hp/free-software-ui.html
Then sit down and answer the question for yourself. Is the feature
good? What's the rationale for it? Answer "why" not just "why not."
Justify in terms of users as a whole, not just users like
yourself. How else can you solve the same problem? etc. If that
leads you to a strong opinion, then please, post the rationale for
discussion to an appropriate bugzilla bug, or to
usability@gnome.org.
Please don't just "me too!" on bugzilla bugs, please don't think
flames will get you anywhere, and please don't repeat rationale
that's already been offered.
Q: Your dumb web pages you made me read talk about solving problems in
fundamental ways instead of adding preferences or workarounds.
What are some examples where metacity has done this?
A: There are quite a few, though many opportunities remain. Sometimes
the real fix involves application changes. The metacity approach is
that it's OK to require apps to change, though there are also
plenty of workarounds in metacity for battles considered too hard
to fight.
Here are some examples:
- fullscreen mode was introduced to allow position constraints,
panel-on-top, and other such things to apply to normal windows
while still allowing video players etc. to "just work"
- "whether to include minimized windows in Alt+Tab" was solved
by putting minimized windows at the *end* of the tab order.
- Whether to pop up a feedback display during Alt+Tab was solved by
having both Alt+Tab and Alt+Esc
- Whether to have a "kill" feature was solved by automatically
detecting and offering to kill stuck apps. Better, metacity
actually does "kill -9" on the process, it doesn't just
disconnect the process from the X server. You'll appreciate this
if you ever did a "kill" on Netscape 4, and watched it keep
eating 100% CPU even though the X server had booted it.
- The workspaces vs. viewports mess was avoided by adding
directional navigation and such to workspaces, see discussion
earlier in this file.
- Instead of configurable placement algorithms, there's just one
that works fairly well most of the time.
- To avoid excess CPU use during opaque move/resize, we rate limit
the updates to the application window's size.
- Instead of configurable "show size of window while resizing,"
it's only shown for windows where it matters, such as terminals.
(Only use-case given for all windows is for web designers
choosing their web browser size, but there are web sites and
desktop backgrounds that do this for you.)
- Using startup notification, applications open on the workspace
where you launched them, not the active workspace when their
window is opened.
- and much more.
Q: I think mutter sucks.
A: Feel free to use any WM you like. The reason metacity follows the
ICCCM and EWMH specifications is that it makes metacity a modular,
interchangeable part in the desktop. libwnck-based apps such as the
GNOME window list will work just fine with any EWMH-compliant WM.
2001-07-25 01:43:41 -04:00
Q: Did you spend a lot of time on this?
A: Originally the answer was no. Sadly the answer is now yes.
2001-07-25 01:43:41 -04:00
Q: How can you claim that you are anti-crack, while still
writing a window manager?
A: I have no comment on that.