I originally put the rebase at the end of view_map, but at this point
the view is still at its native size and will ignore the motion event if
it falls outside of its native size. The only way to do this properly is
to rebase the cursor later - either after sending the configure, after
the view commits with the new size, or after applying the transaction. I
chose to do it after applying the transaction for simplicity.
I then attempted to just call cursor_rebase after applying every
transaction, but this causes crashes when exiting sway (and possibly
other places) because cursor_rebase assumes the tree is in a valid
state.
So my chosen solution introduces transaction_commit_dirty_with_callback
which allows handle_map to register a callback which will run when the
transaction is applied.
The input manager is a singleton object. Passing the sway_input_manager
argument to each of its functions is unnecessary, while removing the
argument makes it obvious to the caller that it's a singleton. This
patch removes the argument and makes the input manager use server.input
instead.
On a similar note:
* sway_input_manager.server is removed in favour of using the server
global.
* seat.input is removed because it can get it from server.input.
Due to a circular dependency, creating seat0 is now done directly in
server_init rather than in input_manager_create. This is because
creating seats must be done after server.input is set.
Lastly, it now stores the default seat name using a constant and removes
a second reference to seat0 (in input_manager_get_default_seat).
This replaces view.using_csd with a new border mode: B_CSD. This also
removes sway_xdg_shell{_v6}_view.deco_mode and
view->has_client_side_decorations as we can now get these from the
border.
You can use `border toggle` to cycle through the modes including CSD, or
use `border csd` to set it directly. The client must support the
xdg-decoration protocol, and the only client I know of that does is the
example in wlroots.
If the client switches from SSD to CSD without us expecting it (via the
server-decoration protocol), we stash the previous border type into
view.saved_border so we can restore it if the client returns to SSD. I
haven't found a way to test this though.
This commit changes the meaning of sway_container so that it only refers
to layout containers and view containers. Workspaces, outputs and the
root are no longer known as containers. Instead, root, outputs,
workspaces and containers are all a type of node, and containers come in
two types: layout containers and view containers.
In addition to the above, this implements type safe variables. This
means we use specific types such as sway_output and sway_workspace
instead of generic containers or nodes. However, it's worth noting that
in a few places places (eg. seat focus and transactions) referring to
them in a generic way is unavoidable which is why we still use nodes in
some places.
If you want a TL;DR, look at node.h, as well as the struct definitions
for root, output, workspace and container. Note that sway_output now
contains a workspaces list, and workspaces now contain a tiling and
floating list, and containers now contain a pointer back to the
workspace.
There are now functions for seat_get_focused_workspace and
seat_get_focused_container. The latter will return NULL if a workspace
itself is focused. Most other seat functions like seat_get_focus and
seat_set_focus now accept and return nodes.
In the config->handler_context struct, current_container has been
replaced with three pointers: node, container and workspace. node is the
same as what current_container was, while workspace is the workspace
that the node resides on and container is the actual container, which
may be NULL if a workspace itself is focused.
The global root_container variable has been replaced with one simply
called root, which is a pointer to the sway_root instance.
The way outputs are created, enabled, disabled and destroyed has
changed. Previously we'd wrap the sway_output in a container when it is
enabled, but as we don't have containers any more it needs a different
approach. The output_create and output_destroy functions previously
created/destroyed the container, but now they create/destroy the
sway_output. There is a new function output_disable to disable an output
without destroying it.
Containers have a new view property. If this is populated then the
container is a view container, otherwise it's a layout container. Like
before, this property is immutable for the life of the container.
Containers have both a `sway_container *parent` and
`sway_workspace *workspace`. As we use specific types now, parent cannot
point to a workspace so it'll be NULL for containers which are direct
children of the workspace. The workspace property is set for all
containers, except those which are hidden in the scratchpad as they have
no workspace.
In some cases we need to refer to workspaces in a container-like way.
For example, workspaces have layout and children, but when using
specific types this makes it difficult. Likewise, it's difficult for a
container to get its parent's layout when the parent could be another
container or a workspace. To make it easier, some helper functions have
been created: container_parent_layout and container_get_siblings.
container_remove_child has been renamed to container_detach and
container_replace_child has been renamed to container_replace.
`container_handle_fullscreen_reparent(con, old_parent)` has had the
old_parent removed. We now unfullscreen the workspace when detaching the
container, so this function is simplified and only needs one argument
now.
container_notify_subtree_changed has been renamed to
container_update_representation. This is more descriptive of its
purpose. I also wanted to be able to call it with whatever container was
changed rather than the container's parent, which makes bubbling up to
the workspace easier.
There are now state structs per node thing. ie. sway_output_state,
sway_workspace_state and sway_container_state.
The focus, move and layout commands have been completely refactored to
work with the specific types. I considered making these a separate PR,
but I'd be backporting my changes only to replace them again, and it's
easier just to test everything at once.
When collecting focus to save into the transaction state, the workspace
needs to look in the tiling list only.
As seat_get_focus_inactive_tiling returns any descendant, the list also
needs to be traversed back up to the direct child of the workspace.
Fixes#2532
When there's multiple transactions in the queue, sway can take a
shortcut by checking if they all operate on the same set of containers.
If they do, it can skip all but the last transaction. The way we tested
for transactions which used the same containers was to exclusive-or
their con IDs together, but this has proved not only to be ineffective
but also has the potential to make sway crash.
This patch replaces the exclusive-or with a loop and container
comparison.
This changes the destroy functions to the following:
* output_begin_destroy
* output_destroy
* workspace_begin_destroy
* workspace_destroy
* container_begin_destroy
* container_destroy
* view_begin_destroy
* view_destroy
The terminology was `destroy` and `free`, and it has been changed to
`begin_destroy` and `destroy` respectively.
When the last output is disconnected, its workspaces will now be stashed
in the root. Upon connection of a new output they will be restored.
There is a new function `workspace_consider_destroy` which decides
whether the given workspace should be destroyed or not (ie. empty and
not visible).
Calling container_begin_destroy will no longer automatically reap the
parents. In some places we want to reap the parents and in some we
don't, so this is left to the caller.
container_reap_empty_recursive and container_reap_empty have been
combined into one function and it will recurse up the tree.
Previously we used a reparent event to detect when a view changes
parent, then sent an output enter/leave to the surfaces if needed. This
worked for tiling views but not floating views, as floating views can
intersect another output without changing parent.
The solution implemented for floating views also applies cleanly to
tiling views, so the previous method has been completely replaced and
the reparent event has been removed.
This introduces a new function container_discover_outputs. This function
compares the container's `current` position to the outputs, sends enter
and leave events as needed, and keeps track of which outputs it's
intersecting in a new `container->outputs` list. If it has entered a new
output with a different scale then the title and marks textures will
also be recreated at the new scale.
The function is called when a transaction applies. This is convenient as
it means we don't have to call it from various places.
There is imperfect rendering when a floating view overlaps two outputs
with different scales. It renders correctly for the most recently
entered output, but there is only one title texture so it renders
incorrectly on the old output.
Fixes#2482
We were removing the saved buffer when one transaction applies, then
didn't have a new buffer to save when the next transaction ran. This
made the rendering code crash as it had no surface to use.
This commit makes it continue to hold the buffer if the view is
destroying and has more transactions. Additionally, a check is added
when saving the buffer to make sure there's no one already there.
Workspaces previously had a magical `workspace->floating` container,
which had a layout of L_FLOATING and whose children were actual floating
views. This allowed some conveniences, but was a hacky solution because
the container has to be exempt from focus, coordinate transactions with
the workspace, and omit emitting IPC events (which we didn't do).
This commit changes it to be a list directly in the sway_workspace. The
L_FLOATING layout is no longer used so this has been removed as well.
* Fixes incorrect check in the swap command (it checked if the
containers had the L_FLOATING layout, but this layout applied to the
magical container).
* Introduces workspace_add_floating
This makes all debug options stored in a single struct rather than in
various places, changes/fixes the behaviour of existing options, and
introduces some new options.
* Fixes damage issues with `-Drender-tree` texture (by removing scissor)
* Offsets the render tree overlay's `y` position for those who have
swaybar at the top
* Replaces `-Ddamage=rerender` with `-Dnodamage`
* Replaces `-Ddamage=highlight` with `-Dhighlight-damage`
* Replaces `-Dtxn-debug` with `-Dtxn-wait`
* Introduces `-Dnoatomic`
* Removes the `create_time` and `ms_arranging` figures from transactions
and the log message. Transactions are created after arranging and the
create time is of no significance.
* Fixes `-Dtxn-debug` (now `-Dtxn-wait`) not working.
This fixes a race condition flicker when unfloating a view which uses
client side decorations.
When the view is floated it has using_csd = true, so the decorations are
not drawn. When unfloating it it changes to false, but this change
wasn't part of transactions so it could potentially render the
decorations around the view while it's waiting for the transaction to
apply.
Commit 4b8e3a885b makes it so only one
transaction is committed (ie. configures sent) at a time. This commit
removes the now-unnecessary code which was used to support concurrent
committed transactions.
* Instead of containers storing a list of instructions which they've
been sent, it now stores a single instruction.
* Containers now have an ntxnrefs property. Previously we knew how many
references there were by the length of the instruction list.
* Instructions no longer need a ready property. It was used to avoid
marking an instruction ready twice when they were in a list, but this is
now avoided because there is only one instruction and we nullify the
container->instruction pointer when it's ready.
* When a transaction applies, we no longer need to consider releasing
and resaving the surface, as we know there are no other committed
transactions.
* transaction_notify_view_ready has been renamed to
view_notify_view_ready_by_serial to make it consistent with
transaction_notify_view_ready_by_size.
* Out-of-memory checks have been added when creating transactions and
instructions.
This fixes an issue where views might commit to a transaction ahead of
the first one, and applying the first transaction causes us to save a
buffer of the wrong size.
This fixes two issues which were both introduced in #2396.
First issue:
The PR changes the location of the buffer save to transaction_apply, but
puts it inside the should_configure block. For unmapping (destroying)
views, should_configure returns false so it wasn't saving the buffer. If
a frame was rendered between the unmap and the transaction applying then
it would result in a crash.
Second issue:
If a destroying view is involved in two transactions, we must not
release the buffer between the transactions because there is no live
buffer to grab any more.
Fixes#2364.
Suppose a view is 600px wide, and we tell it to resize to 601px during a
resize operation. We create a transaction, save the 600px buffer and
send the configure. This buffer is saved into the associated
instruction, and is rendered while we wait for the view to commit a
601px buffer.
Before the view commits the 601px buffer, suppose we tell it to resize
to 602px. The new transaction will also save the buffer, but it's still
the 600px buffer because we haven't received a new one yet.
Then suppose the view commits its original 601px buffer. This completes
the first transaction, so we apply the 601px width to the container.
There's still the second (now only) transaction remaining, so we render
the saved buffer from that. But this is still the 600px buffer, and we
believe it's 601px. Whoops.
The problem here is we can't stack buffers like this. So this commit
removes the saved buffer from the instructions, places it in the view
instead, and re-saves the latest buffer every time the view completes a
transaction and still has further pending transactions.
As saved buffers are now specific to views rather than instructions, the
functions for saving and removing the saved buffer have been moved to
view.c.
The calls to save and restore the buffer have been relocated to more
appropriate functions too, favouring transaction_commit and
transaction_apply rather than transaction_add_container and
transaction_destroy.
The solution used in 073ac425d5 doesn't
work in all cases because the freed instruction might be ahead in the
list, not necessarily behind.
The new solution delays running the queue until after the loop has
finished iterating, thus avoiding the problem completely.
In set_instructions_ready, calling set_instruction_ready may cause any
number of transactions to get applied, which removes them from the list
being iterated. The iteration variables need to be adjusted
accordingly.
When interactively resizing some views (eg. Nautilus), new transactions
are added to the queue faster than the client can process them.
Previously, we would wait for the entire queue to be ready before
applying any of them, but in this case the transactions would time out,
giving the client choppy performance.
This changes the queue handling so it applies the transactions up to the
first waiting transaction, without waiting for the entire queue to be
ready.
Rather than maintain copies of the entire focus stack, this PR
transactionises the focus by introducing two new properties to the
container state and using those when rendering.
* `bool focused` means this container has actual focus. Only one
container should have this equalling true in its current state.
* `struct sway_container *focus_inactive_child` points to the immediate
child that was most recently focused (eg. for tabbed and stacked
containers).
We currently have several ways of setting debug flags, including command
line arguments, environment variables, and compile-time macros. This
replaces the lot with command line flags.
This PR changes the way we handle transactions to a more simple method.
The new method is to mark containers as dirty from low level code
(eg. arranging, or container_destroy, and eventually seat_set_focus),
then call transaction_commit_dirty which picks up those containers and
runs them through a transaction. The old methods of using transactions
(arrange_and_commit, or creating one manually) are now no longer
possible.
The highest-level code (execute_command and view implementation
handlers) will call transaction_commit_dirty, so most other code just
needs to set containers as dirty. This is done by arranging, but can
also be done by calling container_set_dirty.
If we set an instruction as ready twice, it decreases the transaction's
num_waiting a second time and applies the transaction earlier than it
should. This no doubt has undesired effects, probably resulting in a use
after free.
Hopefully fixes the first part of #2207.
A flash of background was happening for two reasons:
1) We were using the xsurface's dimensions to check if the surface is
ready, but these are pending dimensions.
2) In my particular setup, the default geometry of the xsurface does not
intersect any output, which prevented it from receiving a frame done
event. This made the transaction time out and the client would only
redraw once it's been rendered.
