Introduce the API for users (compositors) to create an output from a
head, attach and detach heads, and destroy outputs created this way.
This also adds the backend-facing API to libweston.

In the new API design, a backend creates heads, and the compositor
chooses one or more heads (clone mode) to be driven by an output.
In the future backends will be converted to not create outputs directly
but only in the new create_output hook.

The user subscribes to a heads_changed hook and arranges heads into
outputs from there.

Adding the API this way will allow frontends (main.c) and backends to be
converted one by one. This adds compatiblity paths in
weston_compositor_create_output_with_head() and weston_output_destroy()
so that frontends can be converted first to call these, and then
backends can be converted one by one to the new design. Afterwards, the
compatibility paths will be removed along with weston_output::head.

Currently heads can be added to a disabled output only. This is less
than ideal for clone mode hotplug and should be improved on later.

v4: Remove the wl_output global on head detach if output is enabled.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Add a hook for compositors to get a callback when heads are added or
their connection status changes, to which compositors likely want to
react to by enabling or disabling outputs (API for that to be added
later).

As many head changes as possible should be coalesced into a single
heads_changed call. Therefore the callback is made from an idle task.
This anticipates a future atomic output configuration API, where the
global output configuration is tested and set atomically instead of one
by one.

weston_pending_output_coldplug() needs to manually execute the
heads_changed call so that initial outputs are created before any
plugins get their start-up idle tasks ran. This is especially important
for ivi-shell which does not support output hotplug, and for tests to
guarantee the expected outputs.

v8:
- Change the callback function pointer into a wl_signal. The API is
  changed and renamed.

v6:
- fix a typo
- add comment in weston_pending_output_coldplug()

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v6 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
v6 Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

weston_compositor needs to maintain a list of all available heads, so
that a compositor can pick and choose which heads to take into or out of
use at arbitrary times. The heads may be on or off, and connected or
disconnected.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Enabled is orthogonal from connected. A connected head could be
disabled, or a disconnected head could in the future be enabled.

Compositors quite likely want to check if a head is already enabled
before starting to take it into use.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Heads may be disconnected or connected and the compositor needs to be
able to know the state to know which heads to take into use.

Currently a single head is automatically created with an output, and
outputs are only ever created as connected and destroyed on
disconnection, so it suffices to set connected to true. In the future,
backends are expected to create heads for both connected and
disconnected connectors, so that a connector can be forced on without it
being actually connected.

v6:
- split weston_head_is_enabled() to a new patch

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Heads need to be named, so they can be referenced in logs and
configuration sources.

When clone mode is implemented, output and head names may differ.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

The 'head' member of 'struct weston_output' is going to go unused and
then disappear, so stop using it and find a head from the proper list.

However, this leaves a problem in cms-colord: if you have multiple
monitors driver with the same CRTC, what do you say to the color
management system? The monitors could be different, but all the color
LUTs etc. are in the CRTC and are shared, as is the framebuffer.

Do the simple hack here and just use whatever head happens to be the
first in the list.

The warning is printed in get_output_id(), because if heads are added or
removed while the output is enabled, the id could change.

v6:
- add weston_output_get_first_head(), at first use
- add warning message for nr. heads > 1

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Duplicate these strings to decouple their lifetime from whatever the
backends used. This should prevent hard to catch use after frees and
such problems in the future.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Remove the wl_resource in the head's resource list when we are
removing the wl_output global. We sent global removal events to clients,
the resources should become dummies until clients reap them. Reset user
data so that clients triying to use dummy objects don't hit e.g. a freed
head pointer.

This fixes a theoretical issue: if an enabled output is disabled and
then gets enabled again, mode changes and wl_surface.enter/leave would
still attempt to use the dummy objects. If a client destroyed a dummy
object, we don't have the destructor to remove it from the resource
list, and libweston would hit freed memory.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

The intention is that in the future backends will dynamically allocate
weston_heads based on the resources they have. The lifetime of a
weston_head will be independent of the lifetime of a weston_output it
may be attached to. Backends allocate objects derived from weston_head,
like they currently do for weston_output. Backend will choose when to
destroy a weston_head.

For clone mode, struct weston_output gains head_list member, which is
the list of attached heads that will all show the same framebuffer.
Since heads are growing out of weston_output, management functions are
added.

Detaching a head from an enabled output is allowed to accommodate
disappearing heads. Attaching a head to an enabled output is disallowed
because it may need hardware reconfiguration and testing, and so
requires a weston_output_enable() call.

As a temporary measure, we have one weston_head embedded in
weston_output, so that backends can be migrated individually to the new
allocation scheme.

v8:
- Do not send wp_presentation_feedback.sync_output events for multiple
  wl_output globals in weston_presentation_feedback_present().

v6:
- adapt to upstream changes in weston_output_set_transform()
- use wl_list_for_each_safe in weston_output_release()
- removed weston_output_get_first_head() as it's not needed yet

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
v7 Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

Split out a new function. This is a pure refactoring, no change in
behaviour.

This helps a following patch that adds a loop over output->head_list.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

The user data of a wl_resource representing a wl_output protocol object
used to be a pointer to weston_output. Now that weston_output is being
split, wl_output more accurately refers to weston_head which is a single
monitor.

Change the wl_output user data to point to weston_head.
weston_output_from_resource() is replaced with
weston_head_from_resource().

This change is not strictly necessary, but architecturally it is the
right thing to do. In the future there might appear the need to refer to
a specific head of a cloned pair, for instance.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

As a wl_output represents weston_head, use a weston_head pointer as the
wl_output global's user data.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

The wl_output protocol interface exposes things like monitor make,
model, sub-pixel layout and physical dimensions. Obviously wl_output is
meant to represent a monitor.

The abstraction of a monitor is weston_head. Therefore move the wl_output
global and the bound resources list into weston_head.

When clone mode gets implemented in the future, this means that monitors
driven by the same CRTC will still be represented as separate wl_output
globals. This allows to accurately represent the hardware.

Clone mode that used separate, not frame-locked, CRTCs to drive two
monitors as clones would necessarily also be exposed as separate
wl_output since they have different timings.

v6:
- adapt to upstream changes in weston_output_set_transform()

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

In order to support clone modes, libweston needs the concept of a head
that is separate from weston_output. While weston_output manages buffers
and the repaint state machine, weston_head will represent a single
monitor. In the future it will be possible to have a single
weston_output drive one or more weston_heads for a clone mode that
shares the framebuffers between all cloned heads.

All the fields that are obviously properties of the monitor are moved
from weston_output into weston_head.

As moving the fields requires one to touch all the backends for all the
assingments, introduce setter functions for them while we are here. The
setters are identical to the old assignments, for now.

As a temporary measure, weston_output embeds a single head. Also the
ugly casts in weston_head_set_monitor_strings() will be removed by a
follow-up patch.

Libweston major version is bumped, because weston_output struct layout
is changed.

v7:
- Bump libweston major version.

v6:
- adapt to upstream changes in weston_output_set_transform()

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v5 Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
v6 Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Derek Foreman <derekf@osg.samsung.com>

configure won't work if weston's version is higher than libweston's.

ef57a9b7 added support for window operations such as reporting the
title in escape mode. It implemented this by which-window-op case,
inside the existing which-escape-code case. Whilst it would break out of
the former window-op case, it never broke out of the latter escape-code
case. This would lead to window ops (such as reporting title) falling
through to restoring the saved cursor position.

This doesn't seem at all right, and also fixes a warning with GCC 8.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

This (partially) reverts commit bef76179.
This (partially) reverts commit 4d1cd36c.
This (partially) reverts commit 44fc1be9.
This (partially) reverts commit 6b58ea8c.

The new xwm icon code has proven to be leaky and incomplete, and while
we have patches under consideration to fix the rest of its known problems
they still require changes and review cycles.  Currently the known
leaks have been squashed, but it still picks wrong sized icons and
does no scaling, which can lead to very strange rendering.  At window
close time the wrong sized icon appears above the window during fade out.

This patch reverts the mostly solid bits and keeps the unfinished
bits behind in favor of a simpler revert than removing the whole
thing.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>

This reverts commit 332d1892.
And re-introduces the bug it was intended to fix, see:
https://lists.freedesktop.org/archives/wayland-devel/2017-December/036402.html

Reverting this because it causes harm to all xwayland clients - the
input region no longer gets adjusted when resizing windows.

start an xterm, resize it larger, you can no longer interact with the
new area of the window (including the server side decor).

Hopefully sort the last leaks introduced in commit 6b58ea8c

The window could be destroyed before it had a frame but after it had an icon
(I could trigger this with firefox), and the window could be assigned an icon
twice before it had a frame (I could trigger this with terminology).

The latter leak was
Reported-by: Scott Moreau <oreaus@gmail.com>

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Emre Ucan authored 7 years ago and Pekka Paalanen committed 7 years ago

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Emre Ucan authored 7 years ago and Pekka Paalanen committed 7 years ago

If output_list of compositor is empty, value of
ret is read without initialization.

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Emre Ucan authored 7 years ago and Pekka Paalanen committed 7 years ago

we have to set num_images after import_simple_dmabuf
call. Otherwise, egl_images will not be correctly
referenced in gl_renderer_attach_dmabuf.

(Found by clang source code analyzer)

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Emre Ucan authored 7 years ago and Pekka Paalanen committed 7 years ago

If the constraint is an one-shot constraint, constraint
is freed in disable_pointer_constraint function.
Therefore, we should not try to read freed memory at
"switch (constraint->lifetime)" statement.

The removed code is anyway superfluous. Because
surface destroy signal is only removed, when constraint
is an one-shot constraint.

(Found by clang source code analyzer)

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Emre Ucan authored 7 years ago and Pekka Paalanen committed 7 years ago

assigned values of x, y, ret and layout_surface are
never read.

(Found by clang source code analyzer)

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Emre Ucan authored 7 years ago and Pekka Paalanen committed 7 years ago

ret is overwritten by drmModeAddFB2 call

(Found by clang source code analyzer)

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

A memory leak introduced by 6b58ea8c led to me finding a bigger leak,
which is xwm was calling frame_create() without calling frame_destroy().
This meant that the associated icon_surface was not being destroyed,
leaving the destroy handler for it broken. Here we fix this by calling
frame_destroy() when the window is destroyed and free the reply in
the icon_surface destroy handler.
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

This reverts commit d2cb711d.

I missed the call to cairo_image_surface_create_for_data() which assumes
the data will remain present until the cairo surface is destroyed. It
seems the existence of data depends on the reply not being freed.

This will need a more involved fix.

Sorry, I noticed this just seconds after I pushed the patch.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Scott Moreau authored 7 years ago and Pekka Paalanen committed 7 years ago

Fix memory leak introduced by 6b58ea8c. weston_wm_handle_icon() was
calling xcb_get_property_reply() without freeing the reply.

Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Use stride instead of width for buffer calculation.

[Derek Foreman edited the commit log and removed the leftover
initialization of 'size']

Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

Other backends might want to use it.

Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

Dima Ryazanov authored 7 years ago and Pekka Paalanen committed 7 years ago

"has_discrete" gets set to true in if/else if, but gets left unset otherwise.
So let's initialize it to false.

(This was caught by valgrind.)

Signed-off-by: Dima Ryazanov <dima@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Guido Günther authored 7 years ago and Pekka Paalanen committed 7 years ago

Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Guido Günther authored 7 years ago and Pekka Paalanen committed 7 years ago

Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Guido Günther authored 7 years ago and Pekka Paalanen committed 7 years ago

This allows to check if ZWP_LINUX_BUFFER_PARAMS_V1_FLAGS_Y_INVERT is
interpreted correctly by the compositor.

We introduce an OPT_* bitmask to hold this flag and possible future
command line flags.

Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Guido Günther authored 7 years ago and Pekka Paalanen committed 7 years ago

Got renamed in f9dec679

Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>