this should fix D2036 on the e side by checking validity of an icon
theme once efreet has finished scanning for stuff and if its invalid,
going back to hicolor.
@fix
some odd things happen with auto-clone, so give the user manual
intervention automatically with the randr conf dilaog coming up on an
unconfigured screen being detected.
@fix
this fixes a bizarrre issue that valgrind reported that has the
strings we malloc and fill somehow being freed by xlib. totally didn't
make sense. either way, a workaround here is to strdup them
immediately and free the tmp copy and oddly the bug doesn't happen.
this is very odd.
@fix
this fixes a crash on shutdown when freeing up stuff because an exec
instance wasnt ref++'d and ref--'d when adding an event on the event
queue that pointed to it.
@fix
e backlight was accessing e_comp implicitly during shutdown and comp
is shut down before backlight (that is correct as comp relies on
backlight). this fixes that
@fix
if comp is already shut down, then backlight can't do shutdown things
assuming comp is around, thus handle null e_comp. this fixes segv's
waiting to happen as backlight is shut down after comp and thus e_comp
is a garbage pointer of course
@fix
the values requested by the client will be based on its geometry and not
the geometry of the frame. using the frame geometry here results in windows
moving ↘ based on the top/left frame sizes
fix T2912
since forever, sticky windows have not been allowed to receive focus after
various events, eg. desk flip or window close. in some workflows, however,
this may actually be desired behavior
disabled by default
fix T2837
(2) e.src(s): add keyboard.repeat_delay, keyboard.repeat_rate into e.src files
Summary:
As of now, the default values of repeat delay/rate are being set in e_comp_wl.c.
Those values need to be configurable and will be used in e_comp_wl_init().
The limit of each of the values is defined from -1 to 1000. (maximum 1s).
If one of the two is negative, it means default repeat delay/rate are going to be used.
(e.g. delay:400, rate:25)
Test Plan:
N/A
Signed-off-by: Sung-Jin Park <input.hacker@gmail.com>
Reviewers: raster, stefan_schmidt, gwanglim, devilhorns, zmike
Subscribers: Jeon, ohduna, cedric
Differential Revision: https://phab.enlightenment.org/D3364
Summary:
libwayland-server.so will post an error if the requested version
is higher than the supported one anyway, so there's no point in
doing this.
Using MIN() to pick versions is a client side idiom.
#kansas
Reviewers: zmike, devilhorns
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D3385
We have to use void in a function declaration if we want no function
parameters. Using just empty parenthesis means the function takes an
unspecified number of parameters.
We had it correct for most declarations and this series fixes it for
the rest.
Thanks for the sparse semantic parser for pointing this out.
default values for these is set to a very high value, but it's possible
(and guaranteed) for some clients to set them to a value <= 0, which results
in a broken window
in the case where a client has no geometry set, attempting to update
the client's geometry during frame recalc will guarantee that wrong
geometry is set, resulting in a bad first frame
for security reaons, all the dbus methods that allow you to mess with
e are now in the msgbus module - load at your own risk. this is
irrelevant in x11, but in wayland this matters as wayland is actually
secure.
this also disables restart and shutdown dbus methods still in core.
they are there but non-functional due to possibly being able to be
abused in a wayland universe to "dos attack" the wm.
@fix
due to enabling manual rendering (and als animator frametiem to 10
secons) in e_comp_canvas.c when screensaver is active (blanking is
finished totallly - eg the fade to black) evas weill nto render the
last frame of the animation - skipping it and not rendering another
update until screensaver is disabled. this leaves a subtle ghost of
pixel data which is 1 step before black on the screen (until dpms
turns the monitor off).
this fixes that. this delays enabling manual render for 1 more second
after we have been told the screensaver is active. this is plenty of
time to update all the way to black.
@fix
Summary:
Found this with a quick valgrind session. For paths that are not directories, if we fail the `ecore_file_is_dir` condition memory will not be freed.
The change is to use the default e_path freeing function.
Test Plan:
Run enlightenment in valgrind. It should no longer show a mem leak like below.
```
==6912== 8 bytes in 1 blocks are definitely lost in loss record 186 of 5,940
==6912== at 0x4C28C50: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==6912== by 0x4D8574: e_path_dir_list_get (e_path.c:326)
==6912== by 0x4D1BA4: e_module_init.part.0 (e_module.c:183)
==6912== by 0x4D1DA8: e_module_init (e_module.c:153)
==6912== by 0x4371ED: main (e_main.c:868)
```
Reviewers: zmike
Subscribers: cedric, seoz
Differential Revision: https://phab.enlightenment.org/D3356
This reverts commit bd0c590489.
the m->mirror object can never be deleted except in the case of a client
being removed from the owner desk. removing it at any point will break any
users of the deskmirror api and create other issues
ref T2846
there was previously a frame in which maximized x11 csd clients would
reposition themselves offscreen before rendering
this is now fixed by adjusting the client geometries either upon
unsetting the csd or re-setting it, allowing the client to retain its
expected positioning without jitters
a sticky window previously would always have the desk set for where
it was set as sticky, meaning that anything which tries to access it
will be reading wrong data here.
more useful information to provide is the last desk which the sticky
client was focused on, so update that upon focusing it
in the case that a mouse move event occurs, the compositor should validate
the event to ensure that the mouse cursor is actually over the window that
the event claims to be from
fix T2594
so e has a bit of a problem. we mostly deal with zones, BUt these
zones come from our old xinerama code (this likely should just die
some time) and THIS code gets fed info from e's randr code. we
re-fill/modify as randr finds new screens or things get reconfigured.
thus zones adapt. the problem is now all our zone code really has a
hard time reverse mapping the zone back to where it came from -eg the
randr screen data. you literally can't do a whole bunch of things like
know if that zone was an internal laptop lid or an external screen, or
if it was rotated or even what the dpi is... as you ave no deasy way
to map it back other than by guessing geometry matches.
this fixes that by storing the randr screen id (which should be
unique) fromt he original src randr screen in the xinerama screen and
then in the zone. with this you can do a quick lookup in the e randr
data should you ever need to find the info. this should pave the way
for some other fixes/improvements, but without this they cannot be done.
@fix
previous behavior would result in the nocomp window becoming stuck at a fullscreen
layer when ending nocomp, even if the client was no longer fullscreen
fix T2827
in order to continue rendering an iconic client without breaking icccm,
it's necessary to map the client's window again and unset iconic state
whenever rendering is needed, then re-set states when rendering
stops
ref T2788
Summary:
I noticed this because the description dialog in "Apps > Startup Application"
was never getting properly populated.
Test Plan:
When selecting items in the "Apps > Startup Applications" config the
description should update if a "Comment" field is available in the
.desktop file.
Reviewers: zmike
Subscribers: cedric, seoz
Differential Revision: https://phab.enlightenment.org/D3297
efreet will send change events meaning out desktop file icons are
invalid now. we have refs but content is junk. update these to new
desktop ptr handles by using the exec key to look up new ones on the
change event. this fixes ibar gainign a bunch of blank unknown icons
when any desktop files are touched on the system or user dirs
@fix
wayland clients which have csd must be resized according to window geometry,
not client (surface) geometry. this is somewhat tricky to handle because x11
clients which have csd work the exact opposite way and must continue to be
managed using client geometry
this is not my ideal solution for this issue, but I can't think of a
better one at this time which fully fixes wayland client maximization
due to the (current) synchronous creation of wayland surfaces in efl,
creating these dialogs during startup will result in a crash. the only
way to prevent this and still display the dialogs is to wait a bit and
show the dialogs after init has completed
as promised almost a year ago, this module has been unmaintained to
the point where it cannot function in any way, and attempting to keep
it in the tree is a futile effort
this module can now be found at devs/discomfitor/e_module-access.git
ref 77717523f3
_e_menu_realize was trying to swallow the menu->container_object into
the menu->bg_object, but the menu->bg_object was being created on the
compositor canvas, and the container object was being created on the
e_comp->elm_win.
With recent EFL changes, this causes an abort inside Evas.
To fix this, set the menu->evas to be the Evas from the e_comp->elm_win.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
when applying new csd to a window which already has csd, the previous
csd must be removed in order to apply any new csd offsets in order to
avoid unwanted moving/resizing
man xset => If the `threshold' parameter is provided and 0, the
`acceleration' parameter will be used in the exponent of a more
natural and continuous formula, giving precise control for slow motion
but big reach for fast motion, and a progressive transition for motions
in between. Recommended `acceleration' value in this case is 3/2 to 3,
but not limited to that range.
due to how deferred rendering works, it's possible for a client to
send a second buffer before enlightenment has rendered the first one.
in this situation, it seems that the best solution is to queue successive
buffers (frames) and pop the queue after each render
ref T2784
since ICCCM requires that clients be unmapped while iconified, it's necessary
for the compositor to perform one last render prior to the unmap in order to
ensure that mirror objects will still appear as expected. this render must use
the pixmap buffer data in order to avoid timing issues due to async/deferred
rendering, and it is only necessary for the case of clients rendering with a
native surface
fix T2788
E_Client->comp_data is null after a client has been deleted, so
attempting to handle events which require the dereferencing of this
pointer after a client has been deleted will result in a crash
these events should be rejected after delete regardless, since at this
time the compositor has stopped handling events for the client
ref f42c6aa187
i got a crash here and the bt was broken and i couldnt check if
_e_comp_x_client_data_get() returned null, but it's the only thing
that would make sense, so protect against this to avoid a crash. as
this was a one-off, i can't find out more,
@fix
Summary:
return value is small enough, so there is no risk in casting from unsigned int to signed int.
CID: 1267211
Reviewers: devilhorns, raster, cedric, zmike
Subscribers: seoz, sachin.dev
Differential Revision: https://phab.enlightenment.org/D3179
in the case where a shaped window with many rects exists, there is a high
probability of the damage rect count being huge, leading to massive blocking for
each frame as the compositor attempts to fetch all of these rects from the xserver.
instead, the compositor can shortcut this by forcing a full-window damage any time
the rect count is sufficiently high, trading a blocking socket operation for some
amount of (potential) overdraw.
testing in affected scenarios has shown huge improvements: where previously the entire
compositor would lock up, things work as expected now
see https://bugzilla.mozilla.org/show_bug.cgi?id=1214746 for a sample case
e_config->backlight.idle_dim is actually an unsigned char in the
structure, so use a 0 & 1 for min & max
Signed-off-by: Chris Michael <cp.michael@samsung.com>
it seems that damages for popup windows in some applications,
namely blink-based browsers, are reported incorrectly, resulting
in sometimes having a blank window
so a bit of a config wipe had my comp config reset (some dev
shenanigans) and i got to see what current e "out of box" config is -
and it was horrid for menus. visibilitiy effect was broken. vertial.
default was none - always. forever. like empty. so go back to that. in
fact why do this visibility effect? that was the point of having a
different shadow in the theme - each has its own look + effect. this
makes things far more complex...
Summary:
there is no need to set repeatedly input region even if it's already applied.
and this patch remove the code to clear tiler from client's unmapped case.
this fixes that tiler for input region is removed before applying it to comp object in case client is unmmaped yet.
Reviewers: devilhorns, zmike
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D3076
Sometimes, trying to go to the parent directory wouldn’t work and end up showing some garbage. Unfortunately, the code ended up not NULL-ending the path, which made searching for the path separator buggy.
in some cases, it's possible for a client which expects to render on
the next frame to actually render on the frame after. in these cases,
the compositor must not clear the pixmap image until after the render
has occurred in order to avoid inaccuracies. for this reason, the best
place to flag a client for post-render work is at the time of the client's
render
ref T2762
ref D3120
fixes case where plugging an external monitor would cause the screen
to expand onto that display without the display being effectively usable
until after a restart
the previous methodology was effectively:
attach -> ref(new buffer) x2 / unref(old buffer) x2
...
...
attach -> ref(new buffer) x2 / unref(old buffer) x2
this resulted in buffer management failures and crashing. now the
buffer gets 1x ref before render and 1x unref after render, ensuring
that the lifetime is accurate (assuming evas doesn't lie to us)
now we still have random crashing during resize, but not as much as
before
==24509== Invalid write of size 8
==24509== at 0x502D00: _e_elm_win_trap_del (e_win.c:39)
==24509== by 0x509BFC2: _elm_win_evas_object_smart_del (elm_win.c:1886)
==24509== by 0x91F4643: evas_obj_smart_del (in /usr/lib/libevas.so.1.15.99)
==24509== by 0x91F5B5C: evas_object_smart_del (evas_object_smart.c:1021)
==24509== by 0x91E9107: _evas_object_eo_base_destructor (evas_object_main.c:739)
==24509== by 0xE54A8A3: eo_destructor (in /usr/lib/libeo.so.1.15.99)
==24509== by 0x5086715: _elm_widget_eo_base_destructor (elm_widget.c:5744)
==24509== by 0xE54A8A3: eo_destructor (in /usr/lib/libeo.so.1.15.99)
==24509== by 0xE5443EC: _eo_del_internal (eo_private.h:221)
==24509== by 0xE5443EC: _eo_unref (eo_private.h:295)
==24509== by 0xE5443EC: _eo_do_end (eo.c:546)
==24509== by 0x4D5B1A: _e_obj_dialog_free (e_obj_dialog.c:125)
==24509== by 0x4D61FB: e_object_free (e_object.c:152)
==24509== by 0x4D61FB: e_object_unref (e_object.c:152)
==24509== by 0x4EDC54: _e_sys_logout_after (e_sys.c:750)
==24509== by 0x4ED7AC: _e_sys_action_do (e_sys.c:925)
==24509== by 0x4EE348: e_sys_action_raw_do (e_sys.c:311)
==24509== by 0x4EE43F: _e_sys_comp_done_cb (e_sys.c:66)
==24509== by 0x6097348: _edje_emit_cb (edje_program.c:1476)
==24509== by 0x6097348: _edje_emit_handle (edje_program.c:1405)
==24509== by 0x60924EE: _edje_message_queue_process (edje_message_queue.c:787)
==24509== by 0x60926A6: _edje_job (edje_message_queue.c:154)
==24509== by 0xCC5087A: _ecore_job_event_handler (ecore_job.c:121)
==24509== by 0xCC4B204: _ecore_call_handler_cb (ecore_private.h:390)
==24509== by 0xCC4B204: _ecore_event_call (ecore_events.c:565)
==24509== by 0xCC52AE7: _ecore_main_loop_iterate_internal (ecore_main.c:1927)
==24509== by 0xCC52CD6: ecore_main_loop_begin (ecore_main.c:983)
==24509== by 0x4383F4: main (e_main.c:1047)
==24509== Address 0x14fb1a28 is 1,176 bytes inside a block of size 1,352 free'd
==24509== at 0x4C2A65B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==24509== by 0x4D61FB: e_object_free (e_object.c:152)
==24509== by 0x4D61FB: e_object_unref (e_object.c:152)
==24509== by 0x502CED: _e_elm_win_trap_del (e_win.c:37)
==24509== by 0x509BFC2: _elm_win_evas_object_smart_del (elm_win.c:1886)
==24509== by 0x91F4643: evas_obj_smart_del (in /usr/lib/libevas.so.1.15.99)
==24509== by 0x91F5B5C: evas_object_smart_del (evas_object_smart.c:1021)
==24509== by 0x91E9107: _evas_object_eo_base_destructor (evas_object_main.c:739)
==24509== by 0xE54A8A3: eo_destructor (in /usr/lib/libeo.so.1.15.99)
==24509== by 0x5086715: _elm_widget_eo_base_destructor (elm_widget.c:5744)
==24509== by 0xE54A8A3: eo_destructor (in /usr/lib/libeo.so.1.15.99)
==24509== by 0xE5443EC: _eo_del_internal (eo_private.h:221)
==24509== by 0xE5443EC: _eo_unref (eo_private.h:295)
==24509== by 0xE5443EC: _eo_do_end (eo.c:546)
==24509== by 0x4D5B1A: _e_obj_dialog_free (e_obj_dialog.c:125)
==24509== by 0x4D61FB: e_object_free (e_object.c:152)
==24509== by 0x4D61FB: e_object_unref (e_object.c:152)
==24509== by 0x4EDC54: _e_sys_logout_after (e_sys.c:750)
==24509== by 0x4ED7AC: _e_sys_action_do (e_sys.c:925)
==24509== by 0x4EE348: e_sys_action_raw_do (e_sys.c:311)
==24509== by 0x4EE43F: _e_sys_comp_done_cb (e_sys.c:66)
==24509== by 0x6097348: _edje_emit_cb (edje_program.c:1476)
==24509== by 0x6097348: _edje_emit_handle (edje_program.c:1405)
==24509== by 0x60924EE: _edje_message_queue_process (edje_message_queue.c:787)
==24509== by 0x60926A6: _edje_job (edje_message_queue.c:154)
==24509== by 0xCC5087A: _ecore_job_event_handler (ecore_job.c:121)
==24509== by 0xCC4B204: _ecore_call_handler_cb (ecore_private.h:390)
==24509== by 0xCC4B204: _ecore_event_call (ecore_events.c:565)
==24509== by 0xCC52AE7: _ecore_main_loop_iterate_internal (ecore_main.c:1927)
==24509== by 0xCC52CD6: ecore_main_loop_begin (ecore_main.c:983)
==24509== by 0x4383F4: main (e_main.c:1047)
if this isn't explicitly blocked by config options then allowing resizes
on the unmaximized axes is necessary in order to avoid accidentally
queuing a full unmaximize
according to ICCCM 4.1.4:
Only the client can effect a transition into or out of the Withdrawn state
withdrawn windows cannot be shown under any circumstances. the best that can
be done is to try mapping the window and hope it decides to appear.
to prevent any inadvertent showing of the window before it leaves the
withdrawn state, we play games with the E_Client->ignored flag in order
to skip client evals until we get notified that maybe we want to stop
skipping those evals
ref T2745
gtk apps set an atom which provides information about the area
where non-window content (eg. shadows) may be drawn; this area
must not be used in placement calculations.
the easiest method for implementing this functionality was to add
a case to the compositor geometry interceptors which effectively
flip the client struct geometry values such that the E_Client->client
is outside of the more commonly used E_Client->x/y/w/h
fix T2744
this is not actually supported yet, so behavior of windows using this
feature will be more wayland-like, eg. geometry determined by area
of window+shadow
fix T2744
Summary:
It's more useful for client to bind wl_shm before receiving other global
object's events. Then, App can quickly prepare some buffers. i.e. cursor,
etc.
Signed-off-by: Boram Park <boram1288.park@samsung.com>
Reviewers: stefan_schmidt, gwanglim, raster, zmike, devilhorns
Reviewed By: devilhorns
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D3080
resolves issue where setting a specific kbd would fail to make settings permanent
as well as not propagating the kbd change to the rest of enlightenment
fix T1810
somehow it was possible for client sizes to overflow the zone geometry here
which would end up breaking maximization limits and result in clients
not respecting various geometry boundaries
remove several config vars in advanced performance that frankly don't
work (unless you change them int he dialog). they are not set up on
startup or not even used at all. remove things that don't work anymore!
@fix
this should fix open file handles on unmount by flushing caches first.
not great, but works. long-term have evas not keep file handles open
for 0 refcount cached items.
focusing a client will automatically uniconify and desk flip, so
setting focus on a hidden client should be avoided during eval since
these focus-sets are not "user triggered"
this fixes issues where clients could randomly grab focus from other
desks and also restores expected behavior when restarting e on an
empty vdesk
when using "system" backlight mode, ensure that events are processed
as they occur so that the available backlight level is in sync with
the actual backlight level
fix T2255
due to the presence of color classes on internal windows, e must
ensure that users who set translucent colors for window base objects
get transparency as expected on those parts of the window. it's impossible
to determine in advance whether alpha is needed, so it's best to just
set alpha in all cases
fix T2050
when working with Extremely Serious effects, it may be the case that
a user is rendering at such an advanced level that any attempt by
enlightenment to perform rendering will be like a child trying to
reproduce a masterpiece of art while using fingerpaints
https://www.youtube.com/watch?v=tY6qag5KFx0&hd=1
it's a pretty trivial thing to hand-composite a client, so this will
allow someone to do something like render out a gaussian blur to an fbo
using a client's texture and then render the fbo onto the compositor
canvas with minimal overhead
it's impossible to determine this at the time of calling without adding
some sort of callback here; edje signals are deferred, meaning that
an interested user will not be able to check the state of a client
when it begins to hide
failure to allow pixmaps/clients to be retrived by parent window will
result in api users being greatly inconvenienced after a reparenting has
occurred
ilist widget adds items in a deferred queue, resulting in attempts to
immediately toggle the disabled state having no effect. now there is
a flag so that items added after the disabled flag have been set will
have the correct state
fix T2730
in the case where a window is fullscreen without having the 'fullscreen'
flag set, the previously-used layer must be reapplied upon nocomp end
in order to avoid breaking the compositor
as a result of earlier changes which prevented recursive desk flips,
e_desk_show() now rejects some desk show calls which are invalid such as
a show where the "current" desk does not have the visible flag set. this
behavior is overridden in the case of startup, which is functionally the
same effect as changing the desk count
fix T2717
if windows set to "Always on Top" exist while the option to allow
windows over fullscreen windows is enabled, enabling nocomp will
result in the above windows being stuck over the nocomp window
instead, force the nocomp window to be the top-most window in all cases,
and then put it back if another object appears on the screen over it
fix T2703
when a client is set to "Always on Top", it will be on the same layer
as override clients. this can cause strange stacking and mouse eventing
in cases where these windows occupy the same space and the normal client
is stacked over the override
this prevents an infinite focus loop where focus will be constantly
reapplied between multiple windows if the activated window is not the
refocus window
it seems that since the first version of the enlightenment compositor
in e17, damage events in x11 have never been used correctly. using
the event struct members will only give the bounding box/area instead
of the damaged regions; the real regions must be explicitly fetched
from the server
this removes the need for a lot of hacks which were added over the years
to make override windows render correctly, and also probably reduces
rendering overhead slightly
Before we can use the session_recovery protocol extension we need to register
its global and bind the interface to our implementation.
The callback for the provide_uuid call from the client is just a stub for now.
This reverts commit 466dd8d57d.
sorry - this bug is still there and this tracks references - the
current code DOES NOT DO THIS RIGHT.
No symbol table info available.
No symbol table info available.
out>) at src/bin/e_deskmirror.c:336
mb = 0x1ed7220
#3 0x00007f0e3c864a8d in evas_object_smart_del
(eo_obj=0x8000039ae0101eb8) at lib/evas/canvas/evas_object_smart.c:1016
obj = <optimized out>
sobj = <optimized out>
s = 0x220d540
i = <optimized out>
__FUNCTION__ = "evas_object_smart_del"
(eo_obj=0x8000039ae0101eb8, obj=0x21432e0) at
lib/evas/canvas/evas_object_main.c:739
proxy = <optimized out>
l = <optimized out>
l2 = <optimized out>
texture = <optimized out>
__FUNCTION__ = "_evas_object_eo_base_destructor"
...
and EXACTLY as before p *(mb->m) shows that m is full of garbage.
mb->ec is a garbage pointer, w amd h arte garbage coordinates etc. m
is not tracked. pretty simple. open and close a few windows - mayeb
switch some desktops, THEN restart and see the crashes. its easy
enough to reproduce
if another callback triggered the creation of a deskmirror visual while
the dirty callback was in place, a second mirror object would be created
leading to an orphaned mirror object which retained references to the dm
client and eventually resulting in a crash
the currently visible desk for a zone is stored on the zone struct, so
iterating here is unnecessary. furthermore, at the time when a desk is hidden,
a client may begin receiving mouse events which could trigger a focus-set and
lead to another desk flip. at this time and only this time, the "current" desk
will be marked as not visible, and so this sort of desk show must be rejected
fix T2676
in the case that the canvas window has just had focus set on it, apply this focus
and ensure that no client retains focus
this resolves a race condition where focusing the compositor canvas <-> client
extremely quickly would result in a client trying to steal focus when it was
not actually focused
a notable (but trivial) side effect is that now when flipping desks at high speed while using
mouse-based focus policies, the user is almost guaranteed to end on a desk which
has open windows on it
in the case of recursive desk flips, toggling a desk's visibility may
erroneously send queued evas events to the client's frame object, leading
to a focus-set (mouse-based focus models) which triggers a desk flip
inside the original desk flip. this "inner" desk flip is spurious and
should be ignored
it seems that the reported damage events upon resizing an override window
are not accurate, and so we must force a full damage here while avoiding a
render queue in order to ensure that the full contents of the override will
be rendered in the next frame
fix T2045
this seems to fix an extremely rare issue related to both deskmirror artifacts
and crashes in deskmirror during restart; I was only able to reproduce the crash
twice in the span of over an hour of testing and it seemed to disappear after
this change
This reverts commit 5404adc54f.
I'm strongly against adding refcounts/nullchecks to hide bugs which
are caused by failure to properly track object existence. let's try to avoid
this sort of thing in the future
add deskmirror refs to avoid segv on shutdown/restart due to mb->m
becoming a pointer to freed (now garbage) memory, so trace every
ref/unref and count them to get it right. crash gone!
the shelf theme actually can't know if there is scrollable content in
that direction when a shelf item has autoscroll on and has content in
that direction. this is kind of a missing theme capability that is
incredibly useful if you want to do certain kinds of themes. this is
fairly minor and i'd consider ok for e20 release as we aren't even
doing alphas yet...