The _NET_CLIENT_LIST_STACKING EWMH property is essential for e.g. the tabbar of
chromium to work correctly.
* exwm-input.el: Remove invalid TODO item.
* exwm.el (exwm--init-icccm-ewmh): Add xcb:Atom:_NET_CLIENT_LIST and
xcb:Atom:_NET_CLIENT_LIST_STACKING to _NET_SUPPORTED.
* exwm-layout.el (exwm-layout--refresh): Update _NET_CLIENT_LIST_STACKING.
* exwm-manage.el (exwm-manage--manage-window, exwm-manage--unmanage-window):
Update _NET_CLIENT_LIST.
* exwm-core.el: new buffer-local variable exwm--floating-mode-line-format for
saving mode-line-format when mode-line is hidden
* exwm-floating.el (exwm-floating--fit-frame-to-window)
(exwm-floating-hide-mode-line, exwm-floating-show-mode-line): new functions
for resizing frames, hiding/showing mode-line respectively;
(exwm-floating--set-floating): use exwm-floating--fit-frame-to-window to
resize frames
When Emacs is invoked as `emacsclient -a '' -c`, it creates a frame that can
be deleted without any prompt. This commit removes the `client` parameter from
that frame to avoid such inconvenience.
* When sending the synthetic ConfigureNotify event, make sure we are dealing
with the correct Emacs window
* When managing a floating window, ensure it can be easily pick up by the user
X windows in line-mode receive KeyPress events faked with SendEvent requests
previously. This causes many problems including:
* Some applications (e.g. xterm) ignore synthetic events completely
* KeyPress and KeyRelease evnets arrive disorderly
This commit makes EXWM exploiting AllowEvents requests (in ReplayKeyboard mode)
to forward KeyPress events to X windows instead.
* exwm.el (exwm--on-ClientMessage): Handle fullscreen requests
for frames.
(exwm-init): Initialize workspaces after unlocking events.
* exwm-workspace.el (exwm-workspace--init): Create frames as
invisible, then make them visible only once their OverrideRedirect
property has been set.
* exwm-randr.el (exwm-randr--refresh): New frame parameter
`exwm-geometry'.
* exwm-layout.el (exwm-layout--set-frame-fullscreen): New
function.
The Emacs code is buggy, see https://github.com/ch11ng/exwm/issues/39https://github.com/ch11ng/exwm/pull/42
* exwm-layout.el (exwm-layout-unset-fullscreen)
(exwm-layout-set-fullscreen): Use `user-error' rather than
`cl-assert'.
* exwm-input.el (exwm-input--set-focus): Silently accept unknown
ids. (exwm-input--grab-keyboard) (exwm-input--release-keyboard):
Silently ignore calls for windows that have no buffer.
* exwm-manage.el (exwm-manage--kill-client): Don't throw error
when trying to kill a vanished window.
Some applications (e.g. JNLP) don't set correct values for fields not mentioned
in `value-mask`. This commit corrects this bug together with another Java AWT
specific problem.
* RandR module is now made optional; users can enable it with
`exwm-randr-enable`.
* Correct the calculation of sizes/coordinates at various places.
* Input focus is now tracked with (Emacs) window instead of buffer since the
latter can be ambiguous in multi-screen settings.
This commit should fix most input focus bugs (especially those related to
floating windows). The actual settings of input focus are delayed to exclude
redundant event. Dead code since this commit is removed.
This commit also fixes a bug for non-floating windows converted form floating
state. The workaround for `ido-mode` is also improved to properly handle
`exwm-mode` buffers.
This commit makes a floating window centered to its leading window if it has
a valid WM_TRANSIENT_FOR property set. Other it's placed at the center of the
screen.
* Prevent switching to floating windows or windows on other workspaces
* Provide a workaround for `ido-mode` (can be enabled with
`(exwm-enable-ido-workaround)`)
For some reason, `MappingNotify` events are generated quite frequently and
greatly impact the performance. This commit disables the complete refresh of
keyboard mapping.
Remove `exwm--with-current-id`, which was introduced to as a wrapper to
`with-current-buffer` to do extra checks. Note that in functions run as hooks,
the validation of window ID is still required as they are not synchronized with
events.
`emacsclient` started with `-c` or `-t` argument create a new frame that shall
not be used to manage X windows.
Also fix some related input focus issues (with some remaining unfixed).
Close#17.
With the introduction of ch11ng/xelb@6a7bccc, many weird behaviors should
disappear. These include by not limit to
* race conditions when managing a window (a workaround is provided in @14628a9)
* race conditions when unmanaging a window
This commit removes some corresponding code.
* Relax the conditions to refresh layout; this may introduce some overheads
though
* Fix the problem when `*scratch*` buffer is killed; close#12
* Enhance `exwm-reset` by forcing layout refresh in it. This should allow users
to overcome some layout bugs
* Fix input focus lost after moving window to another workspace
* Enhance `exwm-reset` to provide user a way to reset input focus when it's
lost unexpectedly
* Make sure `exwm-manage--manage-window-queue` is cleaned
* Improve input focus handling after unmanaging a window
* Remove a redundant call to `exwm-layout--show`
Since it takes some time for EXWM to create a buffer for a window (to do some
checking for example), the window may send several MapRequest events before
it's mapped. This commit should fix such issue.
* Fix `exwm-reset`
* Make input mode buffer local
* Allow window to stay in `char-mode` while setting input focus to other window
or switching to other workspace
* Correct ConfigureNotify events sent to fullscreen windows.
* Exit fullscreen mode before switching workspace.
* Temporarily treat `xcb:Atom:_NET_WM_STATE_ABOVE` as
`xcb:Atom:_NET_WM_STATE_FULLSCREEN` since
a) "plugin-container" (Flash Player) seems only set this, and
b) it's not normally used by applications.
This makes fullscreen videos working in e.g. iceweasel.
* Prevent marking the end of a key sequence with a single `C-u'.
* Enable `C-u' prefix for key simulation since it's not possible for users to
define simulation keys starting with `C-u'.
* Make Emacs idle only after the visual parts are updated to prevent from
disturbing users.
* Should use '?\s' instead of '? '.
