mutter/src/core/stack-tracker.c

1393 lines
40 KiB
C
Raw Normal View History

/* -*- mode: C; c-file-style: "gnu"; indent-tabs-mode: nil; -*- */
/*
* stack-tracker:
*
* Track stacking order for compositor
*
* #MetaStackTracker maintains the most accurate view we have at a
* given point of time of the ordering of the children of the root
* window (including override-redirect windows.) This is used to order
* the windows when the compositor draws them.
*
* By contrast, #MetaStack is responsible for keeping track of how we
* think that windows *should* be ordered. For windows we manage
* (non-override-redirect windows), the two stacking orders will be
* the same.
*/
/*
* Copyright (C) 2009 Red Hat, Inc.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 of the
* License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include "config.h"
#include "core/stack-tracker.h"
#include <string.h>
#include "compositor/compositor-private.h"
#include "core/display-private.h"
#include "core/frame.h"
#include "meta/meta-x11-errors.h"
#include "meta/util.h"
#include "x11/meta-x11-display-private.h"
/* The complexity here comes from resolving two competing factors:
*
* - We need to have a view of the stacking order that takes into
* account everything we have done without waiting for events
* back from the X server; we don't want to draw intermediate
* partially-stacked stack states just because we haven't received
* some notification yet.
*
* - Only the X server has an accurate view of the complete stacking;
* when we make a request to restack windows, we don't know how
* it will affect override-redirect windows, because at any point
* applications may restack these windows without our involvement.
*
* The technique we use is that we keep three sets of information:
*
* - The stacking order on the server as known from the last
* event we received.
* - A queue of stacking requests that *we* made subsequent to
* that last event.
* - A predicted stacking order, derived from applying the queued
* requests to the last state from the server.
*
* When we receive a new event: a) we compare the serial in the event to
* the serial of the queued requests and remove any that are now
* no longer pending b) if necessary, drop the predicted stacking
* order to recompute it at the next opportunity.
*
* Possible optimizations:
* Keep the stacks as an array + reverse-mapping hash table to avoid
* linear lookups.
* Keep the stacks as a GList + reverse-mapping hash table to avoid
* linear lookups and to make restacking constant-time.
*/
typedef union _MetaStackOp MetaStackOp;
typedef enum
{
STACK_OP_ADD,
STACK_OP_REMOVE,
STACK_OP_RAISE_ABOVE,
STACK_OP_LOWER_BELOW
} MetaStackOpType;
typedef enum
{
APPLY_DEFAULT = 0,
/* Only do restacking that we can do locally without changing
* the order of X windows. After we've received any stack
* events from the X server, we apply the locally cached
* ops in this mode to handle the non-X parts */
NO_RESTACK_X_WINDOWS = 1 << 0,
/* If the stacking operation wouldn't change the order of X
* windows, ignore it. We use this when applying events received
* from X so that a spontaneous ConfigureNotify (for a move, say)
* doesn't change the stacking of X windows with respect to
* Wayland windows. */
IGNORE_NOOP_X_RESTACK = 1 << 1
} ApplyFlags;
/* MetaStackOp represents a "stacking operation" - a change to
* apply to a window stack. Depending on the context, it could
* either reflect a request we have sent to the server, or a
* notification event we received from the X server.
*/
union _MetaStackOp
{
struct {
MetaStackOpType type;
gulong serial;
guint64 window;
} any;
struct {
MetaStackOpType type;
gulong serial;
guint64 window;
} add;
struct {
MetaStackOpType type;
gulong serial;
guint64 window;
} remove;
struct {
MetaStackOpType type;
gulong serial;
guint64 window;
guint64 sibling;
} raise_above;
struct {
MetaStackOpType type;
gulong serial;
guint64 window;
guint64 sibling;
} lower_below;
};
struct _MetaStackTracker
{
MetaDisplay *display;
MetaStack *stack;
/* This is the serial of the last request we made that was reflected
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
* in xserver_stack
*/
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
gulong xserver_serial;
/* A combined stack containing X and Wayland windows but without
* any unverified operations applied. */
GArray *verified_stack;
/* This is a queue of requests we've made to change the stacking order,
* where we haven't yet gotten a reply back from the server.
*/
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
GQueue *unverified_predictions;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
/* This is how we think the stack is, based on verified_stack, and
* on the unverified_predictions we've made subsequent to
* verified_stack.
*/
GArray *predicted_stack;
/* Idle function used to sync the compositor's view of the window
* stack up with our best guess before a frame is drawn.
*/
guint sync_stack_later;
};
static void
meta_stack_tracker_keep_override_redirect_on_top (MetaStackTracker *tracker);
static inline const char *
get_window_desc (MetaStackTracker *tracker,
guint64 window)
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
{
return meta_display_describe_stack_id (tracker->display, window);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
static void
meta_stack_op_dump (MetaStackTracker *tracker,
MetaStackOp *op,
const char *prefix,
const char *suffix)
{
#ifdef WITH_VERBOSE_MODE
const char *window_desc = get_window_desc (tracker, op->any.window);
#endif
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
switch (op->any.type)
{
case STACK_OP_ADD:
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
meta_topic (META_DEBUG_STACK, "%sADD(%s; %ld)%s",
prefix, window_desc, op->any.serial, suffix);
break;
case STACK_OP_REMOVE:
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
meta_topic (META_DEBUG_STACK, "%sREMOVE(%s; %ld)%s",
prefix, window_desc, op->any.serial, suffix);
break;
case STACK_OP_RAISE_ABOVE:
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
{
meta_topic (META_DEBUG_STACK, "%sRAISE_ABOVE(%s, %s; %ld)%s",
prefix,
window_desc,
get_window_desc (tracker, op->raise_above.sibling),
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
op->any.serial,
suffix);
break;
}
case STACK_OP_LOWER_BELOW:
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
{
meta_topic (META_DEBUG_STACK, "%sLOWER_BELOW(%s, %s; %ld)%s",
prefix,
window_desc,
get_window_desc (tracker, op->lower_below.sibling),
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
op->any.serial,
suffix);
break;
}
}
}
#ifdef WITH_VERBOSE_MODE
static void
stack_dump (MetaStackTracker *tracker,
GArray *stack)
{
guint i;
2014-07-10 16:12:05 -04:00
for (i = 0; i < stack->len; i++)
{
guint64 window = g_array_index (stack, guint64, i);
meta_topic (META_DEBUG_STACK, " %s", get_window_desc (tracker, window));
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
2014-07-10 16:12:05 -04:00
}
#endif /* WITH_VERBOSE_MODE */
2014-07-10 16:12:05 -04:00
static void
meta_stack_tracker_dump (MetaStackTracker *tracker)
{
#ifdef WITH_VERBOSE_MODE
2014-07-10 16:12:05 -04:00
GList *l;
meta_topic (META_DEBUG_STACK, "MetaStackTracker state");
meta_topic (META_DEBUG_STACK, " xserver_serial: %ld", tracker->xserver_serial);
meta_topic (META_DEBUG_STACK, " verified_stack: ");
stack_dump (tracker, tracker->verified_stack);
2014-07-10 16:12:05 -04:00
meta_topic (META_DEBUG_STACK, " unverified_predictions: [");
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
for (l = tracker->unverified_predictions->head; l; l = l->next)
{
MetaStackOp *op = l->data;
meta_stack_op_dump (tracker, op, "", l->next ? ", " : "");
}
meta_topic (META_DEBUG_STACK, "]");
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
if (tracker->predicted_stack)
{
meta_topic (META_DEBUG_STACK, " predicted_stack: ");
stack_dump (tracker, tracker->predicted_stack);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
#endif /* WITH_VERBOSE_MODE */
}
static void
meta_stack_op_free (MetaStackOp *op)
{
g_free (op);
}
static int
find_window (GArray *window_stack,
guint64 window)
{
guint i;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
for (i = 0; i < window_stack->len; i++)
{
guint64 current = g_array_index (window_stack, guint64, i);
if (current == window)
return i;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
return -1;
}
/* Returns TRUE if stack was changed */
static gboolean
move_window_above (GArray *stack,
guint64 window,
int old_pos,
int above_pos,
ApplyFlags apply_flags)
{
int i;
gboolean can_restack_this_window =
(apply_flags & NO_RESTACK_X_WINDOWS) == 0 || !META_STACK_ID_IS_X11 (window);
if (old_pos < above_pos)
{
if ((apply_flags & IGNORE_NOOP_X_RESTACK) != 0)
{
gboolean found_x_window = FALSE;
for (i = old_pos + 1; i <= above_pos; i++)
if (META_STACK_ID_IS_X11 (g_array_index (stack, guint64, i)))
found_x_window = TRUE;
if (!found_x_window)
return FALSE;
}
for (i = old_pos; i < above_pos; i++)
{
if (!can_restack_this_window &&
META_STACK_ID_IS_X11 (g_array_index (stack, guint64, i + 1)))
break;
g_array_index (stack, guint64, i) =
g_array_index (stack, guint64, i + 1);
}
g_array_index (stack, guint64, i) = window;
return i != old_pos;
}
else if (old_pos > above_pos + 1)
{
if ((apply_flags & IGNORE_NOOP_X_RESTACK) != 0)
{
gboolean found_x_window = FALSE;
for (i = above_pos + 1; i < old_pos; i++)
if (META_STACK_ID_IS_X11 (g_array_index (stack, guint64, i)))
found_x_window = TRUE;
if (!found_x_window)
return FALSE;
}
for (i = old_pos; i > above_pos + 1; i--)
{
if (!can_restack_this_window &&
META_STACK_ID_IS_X11 (g_array_index (stack, guint64, i - 1)))
break;
g_array_index (stack, guint64, i) =
g_array_index (stack, guint64, i - 1);
}
g_array_index (stack, guint64, i) = window;
return i != old_pos;
}
else
return FALSE;
}
/* Returns TRUE if stack was changed */
static gboolean
meta_stack_op_apply (MetaStackTracker *tracker,
MetaStackOp *op,
GArray *stack,
ApplyFlags apply_flags)
{
switch (op->any.type)
{
case STACK_OP_ADD:
{
int old_pos;
if (META_STACK_ID_IS_X11 (op->add.window) &&
(apply_flags & NO_RESTACK_X_WINDOWS) != 0)
return FALSE;
old_pos = find_window (stack, op->add.window);
if (old_pos >= 0)
{
meta_topic (META_DEBUG_STACK,
"STACK_OP_ADD: window %s already in stack",
get_window_desc (tracker, op->add.window));
return FALSE;
}
g_array_append_val (stack, op->add.window);
return TRUE;
}
case STACK_OP_REMOVE:
{
int old_pos;
if (META_STACK_ID_IS_X11 (op->remove.window) &&
(apply_flags & NO_RESTACK_X_WINDOWS) != 0)
return FALSE;
old_pos = find_window (stack, op->remove.window);
if (old_pos < 0)
{
meta_topic (META_DEBUG_STACK,
"STACK_OP_REMOVE: window %s not in stack",
get_window_desc (tracker, op->remove.window));
return FALSE;
}
g_array_remove_index (stack, old_pos);
return TRUE;
}
case STACK_OP_RAISE_ABOVE:
{
int old_pos;
int above_pos;
old_pos = find_window (stack, op->raise_above.window);
if (old_pos < 0)
{
meta_topic (META_DEBUG_STACK,
"STACK_OP_RAISE_ABOVE: window %s not in stack",
get_window_desc (tracker, op->raise_above.window));
return FALSE;
}
if (op->raise_above.sibling)
{
above_pos = find_window (stack, op->raise_above.sibling);
if (above_pos < 0)
{
meta_topic (META_DEBUG_STACK,
"STACK_OP_RAISE_ABOVE: sibling window %s not in stack",
get_window_desc (tracker, op->raise_above.sibling));
return FALSE;
}
}
else
{
above_pos = -1;
}
return move_window_above (stack, op->raise_above.window, old_pos, above_pos,
apply_flags);
}
case STACK_OP_LOWER_BELOW:
{
int old_pos;
int above_pos;
old_pos = find_window (stack, op->raise_above.window);
if (old_pos < 0)
{
meta_topic (META_DEBUG_STACK,
"STACK_OP_LOWER_BELOW: window %s not in stack",
get_window_desc (tracker, op->lower_below.window));
return FALSE;
}
if (op->lower_below.sibling)
{
int below_pos;
below_pos = find_window (stack, op->lower_below.sibling);
if (below_pos < 0)
{
meta_topic (META_DEBUG_STACK,
"STACK_OP_LOWER_BELOW: sibling window %s not in stack",
get_window_desc (tracker, op->lower_below.sibling));
return FALSE;
}
above_pos = below_pos - 1;
}
else
{
above_pos = stack->len - 1;
}
return move_window_above (stack, op->lower_below.window, old_pos, above_pos,
apply_flags);
}
}
g_assert_not_reached ();
return FALSE;
}
static GArray *
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
copy_stack (GArray *stack)
{
GArray *copy = g_array_sized_new (FALSE, FALSE, sizeof (guint64), stack->len);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
g_array_set_size (copy, stack->len);
memcpy (copy->data, stack->data, sizeof (guint64) * stack->len);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
return copy;
}
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
static void
query_xserver_stack (MetaDisplay *display,
MetaStackTracker *tracker)
{
MetaX11Display *x11_display = display->x11_display;
Window ignored1, ignored2;
Window *children;
guint n_children;
guint i, old_len;
tracker->xserver_serial = XNextRequest (x11_display->xdisplay);
XQueryTree (x11_display->xdisplay,
x11_display->xroot,
&ignored1, &ignored2, &children, &n_children);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
old_len = tracker->verified_stack->len;
g_array_set_size (tracker->verified_stack, old_len + n_children);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
for (i = 0; i < n_children; i++)
g_array_index (tracker->verified_stack, guint64, old_len + i) = children[i];
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
XFree (children);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
static void
drop_x11_windows (MetaDisplay *display,
MetaStackTracker *tracker)
{
GArray *new_stack;
GList *l;
int i;
tracker->xserver_serial = 0;
new_stack = g_array_new (FALSE, FALSE, sizeof (guint64));
for (i = 0; i < tracker->verified_stack->len; i++)
{
guint64 window = g_array_index (tracker->verified_stack, guint64, i);
if (!META_STACK_ID_IS_X11 (window))
g_array_append_val (new_stack, window);
}
g_array_unref (tracker->verified_stack);
tracker->verified_stack = new_stack;
l = tracker->unverified_predictions->head;
while (l)
{
MetaStackOp *op = l->data;
GList *next = l->next;
if (META_STACK_ID_IS_X11 (op->any.window))
g_queue_remove (tracker->unverified_predictions, op);
l = next;
}
}
static void
on_stack_changed (MetaStack *stack,
MetaStackTracker *tracker)
{
MetaDisplay *display = tracker->display;
GArray *all_root_children_stacked;
GList *l;
GArray *hidden_stack_ids;
GList *sorted;
COGL_TRACE_BEGIN_SCOPED (StackChanged, "Stack changed");
meta_topic (META_DEBUG_STACK, "Syncing window stack to server");
all_root_children_stacked = g_array_new (FALSE, FALSE, sizeof (uint64_t));
hidden_stack_ids = g_array_new (FALSE, FALSE, sizeof (uint64_t));
meta_topic (META_DEBUG_STACK, "Bottom to top: ");
sorted = meta_stack_list_windows (stack, NULL);
for (l = sorted; l; l = l->next)
{
MetaWindow *w = l->data;
uint64_t top_level_window;
uint64_t stack_id;
if (w->unmanaging)
continue;
meta_topic (META_DEBUG_STACK, " %u:%d - %s ",
w->layer, w->stack_position, w->desc);
if (w->frame)
top_level_window = w->frame->xwindow;
else
top_level_window = w->xwindow;
if (w->client_type == META_WINDOW_CLIENT_TYPE_X11)
stack_id = top_level_window;
else
stack_id = w->stamp;
/* We don't restack hidden windows along with the rest, though they are
* reflected in the _NET hints. Hidden windows all get pushed below
* the screens fullscreen guard_window. */
if (w->hidden)
{
g_array_append_val (hidden_stack_ids, stack_id);
continue;
}
g_array_append_val (all_root_children_stacked, stack_id);
}
if (display->x11_display)
{
uint64_t guard_window_id;
/* The screen guard window sits above all hidden windows and acts as
* a barrier to input reaching these windows. */
guard_window_id = display->x11_display->guard_window;
g_array_append_val (hidden_stack_ids, guard_window_id);
}
/* Sync to server */
meta_topic (META_DEBUG_STACK, "Restacking %u windows",
all_root_children_stacked->len);
meta_stack_tracker_restack_managed (tracker,
(uint64_t *)all_root_children_stacked->data,
all_root_children_stacked->len);
meta_stack_tracker_restack_at_bottom (tracker,
(uint64_t *)hidden_stack_ids->data,
hidden_stack_ids->len);
g_array_free (hidden_stack_ids, TRUE);
g_array_free (all_root_children_stacked, TRUE);
g_list_free (sorted);
}
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
MetaStackTracker *
meta_stack_tracker_new (MetaStack *stack)
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
{
MetaStackTracker *tracker;
tracker = g_new0 (MetaStackTracker, 1);
tracker->display = stack->display;
tracker->stack = stack;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
tracker->verified_stack = g_array_new (FALSE, FALSE, sizeof (guint64));
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
tracker->unverified_predictions = g_queue_new ();
g_signal_connect (tracker->display,
"x11-display-setup",
G_CALLBACK (query_xserver_stack),
tracker);
g_signal_connect (tracker->display,
"x11-display-closing",
G_CALLBACK (drop_x11_windows),
tracker);
g_signal_connect (tracker->stack, "changed",
G_CALLBACK (on_stack_changed),
tracker);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
meta_stack_tracker_dump (tracker);
return tracker;
}
void
meta_stack_tracker_free (MetaStackTracker *tracker)
{
if (tracker->sync_stack_later)
{
MetaCompositor *compositor =
meta_display_get_compositor (tracker->display);
MetaLaters *laters = meta_compositor_get_laters (compositor);
meta_laters_remove (laters, tracker->sync_stack_later);
}
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
g_array_free (tracker->verified_stack, TRUE);
if (tracker->predicted_stack)
g_array_free (tracker->predicted_stack, TRUE);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
g_queue_foreach (tracker->unverified_predictions, (GFunc)meta_stack_op_free, NULL);
g_queue_free (tracker->unverified_predictions);
tracker->unverified_predictions = NULL;
g_signal_handlers_disconnect_by_func (tracker->display,
(gpointer)query_xserver_stack,
tracker);
g_signal_handlers_disconnect_by_func (tracker->display,
drop_x11_windows,
tracker);
g_signal_handlers_disconnect_by_func (tracker->stack,
on_stack_changed,
tracker);
g_free (tracker);
}
static void
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
stack_tracker_apply_prediction (MetaStackTracker *tracker,
MetaStackOp *op)
{
gboolean free_at_end = FALSE;
/* If this operation doesn't involve restacking X windows then it's
* implicitly verified. We can apply it immediately unless there
* are outstanding X restacks that haven't yet been confirmed.
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
*/
if (op->any.serial == 0 &&
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
tracker->unverified_predictions->length == 0)
{
if (meta_stack_op_apply (tracker, op, tracker->verified_stack, APPLY_DEFAULT))
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
meta_stack_tracker_queue_sync_stack (tracker);
free_at_end = TRUE;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
else
{
meta_stack_op_dump (tracker, op, "Predicting: ", "");
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
g_queue_push_tail (tracker->unverified_predictions, op);
}
if (!tracker->predicted_stack ||
meta_stack_op_apply (tracker, op, tracker->predicted_stack, APPLY_DEFAULT))
meta_stack_tracker_queue_sync_stack (tracker);
if (free_at_end)
meta_stack_op_free (op);
meta_stack_tracker_dump (tracker);
}
void
meta_stack_tracker_record_add (MetaStackTracker *tracker,
guint64 window,
gulong serial)
{
MetaStackOp *op = g_new0 (MetaStackOp, 1);
op->any.type = STACK_OP_ADD;
op->any.serial = serial;
op->any.window = window;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
stack_tracker_apply_prediction (tracker, op);
}
void
meta_stack_tracker_record_remove (MetaStackTracker *tracker,
guint64 window,
gulong serial)
{
MetaStackOp *op = g_new0 (MetaStackOp, 1);
op->any.type = STACK_OP_REMOVE;
op->any.serial = serial;
op->any.window = window;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
stack_tracker_apply_prediction (tracker, op);
}
static void
meta_stack_tracker_record_raise_above (MetaStackTracker *tracker,
guint64 window,
guint64 sibling,
gulong serial)
{
MetaStackOp *op = g_new0 (MetaStackOp, 1);
op->any.type = STACK_OP_RAISE_ABOVE;
op->any.serial = serial;
op->any.window = window;
op->raise_above.sibling = sibling;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
stack_tracker_apply_prediction (tracker, op);
}
static void
meta_stack_tracker_record_lower_below (MetaStackTracker *tracker,
guint64 window,
guint64 sibling,
gulong serial)
{
MetaStackOp *op = g_new0 (MetaStackOp, 1);
op->any.type = STACK_OP_LOWER_BELOW;
op->any.serial = serial;
op->any.window = window;
op->lower_below.sibling = sibling;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
stack_tracker_apply_prediction (tracker, op);
}
static void
stack_tracker_event_received (MetaStackTracker *tracker,
MetaStackOp *op)
{
gboolean need_sync = FALSE;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
/* If the event is older than our initial query, then it's
* already included in our tree. Just ignore it. */
if (op->any.serial < tracker->xserver_serial)
return;
meta_stack_op_dump (tracker, op, "Stack op event received: ", "");
/* First we apply any operations that we have queued up that depended
* on X operations *older* than what we received .. those operations
* must have been ignored by the X server, so we just apply the
* operations we have as best as possible while not moving windows.
*/
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
while (tracker->unverified_predictions->head)
{
MetaStackOp *queued_op = tracker->unverified_predictions->head->data;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
if (queued_op->any.serial >= op->any.serial)
break;
meta_stack_op_apply (tracker, queued_op, tracker->verified_stack,
NO_RESTACK_X_WINDOWS);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
g_queue_pop_head (tracker->unverified_predictions);
meta_stack_op_free (queued_op);
need_sync = TRUE;
}
/* Then we apply the received event. If it's a spontaneous event
* based on stacking we didn't trigger, this is the only handling. If we
* triggered it, we do the X restacking here, and then any residual
* local-only Wayland stacking below.
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
*/
if (meta_stack_op_apply (tracker, op, tracker->verified_stack,
IGNORE_NOOP_X_RESTACK))
need_sync = TRUE;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
/* What is left to process is the prediction corresponding to the event
* (if any), and then any subsequent Wayland-only events we can just
* go ahead and do now.
*/
while (tracker->unverified_predictions->head)
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
{
MetaStackOp *queued_op = tracker->unverified_predictions->head->data;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
if (queued_op->any.serial > op->any.serial)
break;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
meta_stack_op_apply (tracker, queued_op, tracker->verified_stack,
NO_RESTACK_X_WINDOWS);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
g_queue_pop_head (tracker->unverified_predictions);
meta_stack_op_free (queued_op);
need_sync = TRUE;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
}
if (need_sync)
{
if (tracker->predicted_stack)
{
g_array_free (tracker->predicted_stack, TRUE);
tracker->predicted_stack = NULL;
}
meta_stack_tracker_queue_sync_stack (tracker);
}
meta_stack_tracker_dump (tracker);
}
void
meta_stack_tracker_create_event (MetaStackTracker *tracker,
XCreateWindowEvent *event)
{
MetaStackOp op;
op.any.type = STACK_OP_ADD;
op.any.serial = event->serial;
op.add.window = event->window;
stack_tracker_event_received (tracker, &op);
}
void
meta_stack_tracker_destroy_event (MetaStackTracker *tracker,
XDestroyWindowEvent *event)
{
MetaStackOp op;
op.any.type = STACK_OP_REMOVE;
op.any.serial = event->serial;
op.remove.window = event->window;
stack_tracker_event_received (tracker, &op);
}
void
meta_stack_tracker_reparent_event (MetaStackTracker *tracker,
XReparentEvent *event)
{
if (event->parent == event->event)
{
MetaStackOp op;
op.any.type = STACK_OP_ADD;
op.any.serial = event->serial;
op.add.window = event->window;
stack_tracker_event_received (tracker, &op);
}
else
{
MetaStackOp op;
op.any.type = STACK_OP_REMOVE;
op.any.serial = event->serial;
op.remove.window = event->window;
stack_tracker_event_received (tracker, &op);
}
}
void
meta_stack_tracker_configure_event (MetaStackTracker *tracker,
XConfigureEvent *event)
{
MetaStackOp op;
op.any.type = STACK_OP_RAISE_ABOVE;
op.any.serial = event->serial;
op.raise_above.window = event->window;
op.raise_above.sibling = event->above;
stack_tracker_event_received (tracker, &op);
}
static gboolean
meta_stack_tracker_is_guard_window (MetaStackTracker *tracker,
uint64_t stack_id)
{
MetaX11Display *x11_display = tracker->display->x11_display;
if (!x11_display)
return FALSE;
return stack_id == x11_display->guard_window;
}
/**
* meta_stack_tracker_get_stack:
* @tracker: a #MetaStackTracker
* @windows: location to store list of windows, or %NULL
* @n_windows: location to store count of windows, or %NULL
*
* @windows will contain the most current view we have of the stacking order
* of the children of the root window. The returned array contains
* everything: InputOnly windows, override-redirect windows,
* hidden windows, etc. Some of these will correspond to MetaWindow
* objects, others won't.
*
* Assuming that no other clients have made requests that change
* the stacking order since we last received a notification, the
* returned list of windows is exactly that you'd get as the
* children when calling XQueryTree() on the root window.
*/
void
meta_stack_tracker_get_stack (MetaStackTracker *tracker,
guint64 **windows,
int *n_windows)
{
GArray *stack;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
if (tracker->unverified_predictions->length == 0)
{
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
stack = tracker->verified_stack;
}
else
{
if (tracker->predicted_stack == NULL)
{
GList *l;
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
tracker->predicted_stack = copy_stack (tracker->verified_stack);
for (l = tracker->unverified_predictions->head; l; l = l->next)
{
MetaStackOp *op = l->data;
meta_stack_op_apply (tracker, op, tracker->predicted_stack, APPLY_DEFAULT);
}
}
stack = tracker->predicted_stack;
}
if (windows)
*windows = (guint64 *)stack->data;
if (n_windows)
*n_windows = stack->len;
}
/**
* meta_stack_tracker_sync_stack:
* @tracker: a #MetaStackTracker
*
* Informs the compositor of the current stacking order of windows,
* based on the predicted view maintained by the #MetaStackTracker.
*/
void
meta_stack_tracker_sync_stack (MetaStackTracker *tracker)
{
guint64 *windows;
GList *meta_windows;
int n_windows;
int i;
if (tracker->sync_stack_later)
{
MetaLaters *laters;
laters = meta_compositor_get_laters (tracker->display->compositor);
meta_laters_remove (laters, tracker->sync_stack_later);
tracker->sync_stack_later = 0;
}
meta_stack_tracker_keep_override_redirect_on_top (tracker);
meta_stack_tracker_get_stack (tracker, &windows, &n_windows);
meta_windows = NULL;
for (i = 0; i < n_windows; i++)
{
guint64 window = windows[i];
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
if (META_STACK_ID_IS_X11 (window))
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
{
MetaX11Display *x11_display = tracker->display->x11_display;
MetaWindow *meta_window = NULL;
if (x11_display)
meta_window = meta_x11_display_lookup_x_window (x11_display, (Window) window);
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
/* When mapping back from xwindow to MetaWindow we have to be a bit careful;
* children of the root could include unmapped windows created by toolkits
* for internal purposes, including ones that we have registered in our
* XID => window table. (Wine uses a toplevel for _NET_WM_USER_TIME_WINDOW;
* see window-prop.c:reload_net_wm_user_time_window() for registration.)
*/
if (meta_window &&
((Window)window == meta_window->xwindow ||
(meta_window->frame && (Window)window == meta_window->frame->xwindow)))
Add support for stacking X and Wayland windows together This breaks down the assumptions in stack-tracker.c and stack.c that Mutter is only stacking X windows. The stack tracker now tracks windows using a MetaStackWindow structure which is a union with a type member so that X windows can be distinguished from Wayland windows. Some notable changes are: Queued stack tracker operations that affect Wayland windows will not be associated with an X serial number. If an operation only affects a Wayland window and there are no queued stack tracker operations ("unvalidated predictions") then the operation is applied immediately since there is no server involved with changing the stacking for Wayland windows. The stack tracker can no longer respond to X events by turning them into stack operations and discarding the predicted operations made prior to that event because operations based on X events don't know anything about the stacking of Wayland windows. Instead of discarding old predictions the new approach is to trust the predictions but whenever we receive an event from the server that affects stacking we cross-reference with the predicted stack and check for consistency. So e.g. if we have an event that says ADD window A then we apply the predictions (up to the serial for that event) and verify the predicted state includes a window A. Similarly if an event says RAISE_ABOVE(B, C) we can apply the predictions (up to the serial for that event) and verify that window B is above C. If we ever receive spurious stacking events (with a serial older than we would expect) or find an inconsistency (some things aren't possible to predict from the compositor) then we hit a re-synchronization code-path that will query the X server for the full stacking order and then use that stack to walk through our combined stack and force the X windows to match the just queried stack but avoiding disrupting the relative stacking of Wayland windows. This will be relatively expensive but shouldn't be hit for compositor initiated restacking operations where our predictions should be accurate. The code in core/stack.c that deals with synchronizing the window stack with the X server had to be updated quite heavily. In general the patch avoids changing the fundamental approach being used but most of the code did need some amount of re-factoring to consider what re-stacking operations actually involve X or not and when we need to restack X windows we sometimes need to search for a suitable X sibling to restack relative too since the closest siblings may be Wayland windows.
2012-04-05 06:22:13 -04:00
meta_windows = g_list_prepend (meta_windows, meta_window);
}
else
meta_windows = g_list_prepend (meta_windows,
meta_display_lookup_stamp (tracker->display, window));
}
meta_compositor_sync_stack (tracker->display->compositor,
meta_windows);
g_list_free (meta_windows);
meta_display_restacked (tracker->display);
}
static gboolean
stack_tracker_sync_stack_later (gpointer data)
{
meta_stack_tracker_sync_stack (data);
return FALSE;
}
/**
* meta_stack_tracker_queue_sync_stack:
* @tracker: a #MetaStackTracker
*
* Queue informing the compositor of the new stacking order before the
* next redraw. (See meta_stack_tracker_sync_stack()). This is called
* internally when the stack of X windows changes, but also needs be
* called directly when we an undecorated window is first shown or
* withdrawn since the compositor's stacking order (which contains only
* the windows that have a corresponding MetaWindow) will change without
* any change to the stacking order of the X windows, if we are creating
* or destroying MetaWindows.
*/
void
meta_stack_tracker_queue_sync_stack (MetaStackTracker *tracker)
{
MetaLaters *laters;
if (tracker->sync_stack_later != 0)
return;
laters = meta_compositor_get_laters (tracker->display->compositor);
tracker->sync_stack_later = meta_laters_add (laters, META_LATER_SYNC_STACK,
stack_tracker_sync_stack_later,
tracker, NULL);
}
/* When moving an X window we sometimes need an X based sibling.
*
* If the given sibling is X based this function returns it back
* otherwise it searches downwards looking for the nearest X window.
*
* If no X based sibling could be found return NULL. */
static Window
find_x11_sibling_downwards (MetaStackTracker *tracker,
guint64 sibling)
{
guint64 *windows;
int n_windows;
int i;
if (META_STACK_ID_IS_X11 (sibling))
return (Window)sibling;
meta_stack_tracker_get_stack (tracker,
&windows, &n_windows);
/* NB: Children are in order from bottom to top and we
* want to search downwards for the nearest X window.
*/
for (i = n_windows - 1; i >= 0; i--)
if (windows[i] == sibling)
break;
for (; i >= 0; i--)
{
if (META_STACK_ID_IS_X11 (windows[i]))
return (Window)windows[i];
}
return None;
}
static Window
find_x11_sibling_upwards (MetaStackTracker *tracker,
guint64 sibling)
{
guint64 *windows;
int n_windows;
int i;
if (META_STACK_ID_IS_X11 (sibling))
return (Window)sibling;
meta_stack_tracker_get_stack (tracker,
&windows, &n_windows);
for (i = 0; i < n_windows; i++)
if (windows[i] == sibling)
break;
for (; i < n_windows; i++)
{
if (META_STACK_ID_IS_X11 (windows[i]))
return (Window)windows[i];
}
return None;
}
static void
meta_stack_tracker_lower_below (MetaStackTracker *tracker,
guint64 window,
guint64 sibling)
{
gulong serial = 0;
MetaX11Display *x11_display = tracker->display->x11_display;
if (META_STACK_ID_IS_X11 (window))
{
XWindowChanges changes;
changes.sibling = sibling ? find_x11_sibling_upwards (tracker, sibling) : None;
if (changes.sibling != find_x11_sibling_upwards (tracker, window))
{
serial = XNextRequest (x11_display->xdisplay);
meta_x11_error_trap_push (x11_display);
changes.stack_mode = changes.sibling ? Below : Above;
XConfigureWindow (x11_display->xdisplay,
window,
(changes.sibling ? CWSibling : 0) | CWStackMode,
&changes);
meta_x11_error_trap_pop (x11_display);
}
}
meta_stack_tracker_record_lower_below (tracker,
window, sibling,
serial);
}
static void
meta_stack_tracker_raise_above (MetaStackTracker *tracker,
guint64 window,
guint64 sibling)
{
gulong serial = 0;
MetaX11Display *x11_display = tracker->display->x11_display;
if (META_STACK_ID_IS_X11 (window))
{
XWindowChanges changes;
changes.sibling = sibling ? find_x11_sibling_downwards (tracker, sibling) : None;
if (changes.sibling != find_x11_sibling_downwards (tracker, window))
{
serial = XNextRequest (x11_display->xdisplay);
meta_x11_error_trap_push (x11_display);
changes.stack_mode = changes.sibling ? Above : Below;
XConfigureWindow (x11_display->xdisplay,
(Window)window,
(changes.sibling ? CWSibling : 0) | CWStackMode,
&changes);
meta_x11_error_trap_pop (x11_display);
}
}
meta_stack_tracker_record_raise_above (tracker, window,
sibling, serial);
}
void
meta_stack_tracker_lower (MetaStackTracker *tracker,
guint64 window)
{
meta_stack_tracker_raise_above (tracker, window, None);
}
static void
meta_stack_tracker_keep_override_redirect_on_top (MetaStackTracker *tracker)
{
guint64 *stack;
int n_windows, i;
int topmost_non_or;
meta_stack_tracker_get_stack (tracker, &stack, &n_windows);
for (i = n_windows - 1; i >= 0; i--)
{
MetaWindow *window;
window = meta_display_lookup_stack_id (tracker->display, stack[i]);
if (window && window->layer != META_LAYER_OVERRIDE_REDIRECT)
break;
}
topmost_non_or = i;
for (i -= 1; i >= 0; i--)
{
MetaWindow *window;
if (meta_stack_tracker_is_guard_window (tracker, stack[i]))
break;
window = meta_display_lookup_stack_id (tracker->display, stack[i]);
if (window && window->layer == META_LAYER_OVERRIDE_REDIRECT)
{
meta_stack_tracker_raise_above (tracker, stack[i], stack[topmost_non_or]);
meta_stack_tracker_get_stack (tracker, &stack, &n_windows);
topmost_non_or -= 1;
}
}
}
void
meta_stack_tracker_restack_managed (MetaStackTracker *tracker,
const guint64 *managed,
int n_managed)
{
guint64 *windows;
int n_windows;
int old_pos, new_pos;
COGL_TRACE_BEGIN_SCOPED (StackTrackerRestackManaged,
"StackTracker: Restack Managed");
if (n_managed == 0)
return;
COGL_TRACE_BEGIN_SCOPED (StackTrackerRestackManagedGet,
"StackTracker: Restack Managed (get)");
meta_stack_tracker_get_stack (tracker, &windows, &n_windows);
/* If the top window has to be restacked, we don't want to move it to the very
* top of the stack, since apps expect override-redirect windows to stay near
* the top of the X stack; we instead move it above all managed windows (or
* above the guard window if there are no non-hidden managed windows.)
*/
old_pos = n_windows - 1;
for (old_pos = n_windows - 1; old_pos >= 0; old_pos--)
{
MetaWindow *old_window = meta_display_lookup_stack_id (tracker->display, windows[old_pos]);
if ((old_window && !old_window->override_redirect && !old_window->unmanaging) ||
meta_stack_tracker_is_guard_window (tracker, windows[old_pos]))
break;
}
g_assert (old_pos >= 0);
COGL_TRACE_END (StackTrackerRestackManagedGet);
COGL_TRACE_BEGIN_SCOPED (StackTrackerRestackManagedRaise,
"StackTracker: Restack Managed (raise)");
new_pos = n_managed - 1;
if (managed[new_pos] != windows[old_pos])
{
/* Move the first managed window in the new stack above all managed windows */
meta_stack_tracker_raise_above (tracker, managed[new_pos], windows[old_pos]);
meta_stack_tracker_get_stack (tracker, &windows, &n_windows);
/* Moving managed[new_pos] above windows[old_pos], moves the window at old_pos down by one */
}
COGL_TRACE_END (StackTrackerRestackManagedRaise);
old_pos--;
new_pos--;
COGL_TRACE_BEGIN_SCOPED (StackTrackerRestackManagedRestack,
"StackTracker: Restack Managed (restack)");
while (old_pos >= 0 && new_pos >= 0)
{
if (meta_stack_tracker_is_guard_window (tracker, windows[old_pos]))
break;
if (windows[old_pos] == managed[new_pos])
{
old_pos--;
new_pos--;
continue;
}
MetaWindow *old_window = meta_display_lookup_stack_id (tracker->display, windows[old_pos]);
if (!old_window || old_window->override_redirect || old_window->unmanaging)
{
old_pos--;
continue;
}
meta_stack_tracker_lower_below (tracker, managed[new_pos], managed[new_pos + 1]);
meta_stack_tracker_get_stack (tracker, &windows, &n_windows);
/* Moving managed[new_pos] above windows[old_pos] moves the window at old_pos down by one,
* we'll examine it again to see if it matches the next new window */
old_pos--;
new_pos--;
}
COGL_TRACE_END (StackTrackerRestackManagedRestack);
COGL_TRACE_BEGIN_SCOPED (StackTrackerRestackManagedLower,
"StackTracker: Restack Managed (lower)");
while (new_pos > 0)
{
meta_stack_tracker_lower_below (tracker, managed[new_pos], managed[new_pos - 1]);
new_pos--;
}
COGL_TRACE_END (StackTrackerRestackManagedLower);
}
void
meta_stack_tracker_restack_at_bottom (MetaStackTracker *tracker,
const guint64 *new_order,
int n_new_order)
{
guint64 *windows;
int n_windows;
int pos;
COGL_TRACE_BEGIN_SCOPED (StackTrackerRestackAtBottom,
"Stack tracker: Restack at bottom");
meta_stack_tracker_get_stack (tracker, &windows, &n_windows);
for (pos = 0; pos < n_new_order; pos++)
{
if (pos >= n_windows || windows[pos] != new_order[pos])
{
if (pos == 0)
meta_stack_tracker_lower (tracker, new_order[pos]);
else
meta_stack_tracker_raise_above (tracker, new_order[pos], new_order[pos - 1]);
meta_stack_tracker_get_stack (tracker, &windows, &n_windows);
}
}
}