Generally we never want to lock a buffer for write access if it is at
the "head" on the surfaceflinger side. The only exception (1) is when
the buffer is not currently in use AND there is at least one queued
buffer -- in which case, SurfaceFlinger will never use said buffer
anymore, because on the next composition around, it will be able to
retire the first queued buffer.
The logic above relies on SurfaceFlinger always retiring
and locking a buffer before composition -- unfortunately this
didn't happen during a screenshot.
This could leave us in a situation where a buffer is locked by the
application for write, and used by SurfaceFlinger for texturing,
causing a hang.
Here, we fix this issue by never assuming the exception (1), it was
intended as an optimization allowing ANativeWindow::lockBuffer() to
return sooner and was justified when most of SF composition was
done in software. The actual buffer locking is now ensured by
gralloc. We could have handled screenshots in a similar way to
a regular composition, but it could have caused glitches on screen,
essentially, taking a screenshot could cause to skip a frame.
now that we removed the notion of a "inUse" buffer in surfaceflinger
a lot of code can be simplified / removed.
noteworthy, the whole concept of "unlockClient" wrt. "compositionComplete"
is also gone.
We used to guarantee that a layer in SurfaceFlinger would never be
destroyed before all references (to its ISurface) on the client
side would be released. At some point, this guarantee got
relaxed to allow to free gralloc resources sooner. This last
change was incorrect, because:
- in implementations with reference-counting the gralloc resources
wouldn't be released anyways, until all the mapping were gone
- in implementations without ref counting, the client side
would most likely crash or do something bad
- it also caused the SharedBufferStack slot to be reallocated
to another surface, which could be problematic if the client
continued to use the surface after the window manager destroyed it.
So, we essentially reinstate the guarantee that layers won't be
destroyed until after all references to their ISurface are
released.
NOTE: This doesn't entirely fix 3306150 because there is another
problem there where the Browser continues to use a surface after it
has been destroyed.
also improve SurfaceFlinger 'dumpsys' log
list the purgatory, which shows windows that have been closed,
but for which the client still has references.
InputReader::getSwitchState always returns AKEY_STATE_UNKNOWN
because SwitchInputMapper::getSources() returns 0 which cannot
match any source mask including AINPUT_SOURCE_ANY. As a result
initial lid switch detection is broken.
This change adds a new source constant AINPUT_SOURCE_SWITCH
that indicates that the source has switches.
Change-Id: I5321ecf0ce84f1c2b4535f6c163d3f4dcf9b7a9b
EGLUtils::selectConfigForPixelFormat() now uses EGL_NATIVE_VISUAL_ID
to select a config with the proper format. this is more robust and
future proof.
Change-Id: Ib85e0974160945d838956b50a3cac4de78618957
Two of the cases in the ISurfaceComposer onTransact switch statement
were missing 'break' statements at the end, and would fall through to
the next case block. This change adds those break statements.
Change-Id: I6dcc84263d3ea03d94612c667103283846b3dee1
Adds a new virtualKeyQuietTimeMillis configuration resource that sets
the duration for which virtual keys will be dropped after recent touches
on screen. The default value is 0; it is intended to be overridden
per device using a resource overlay.
This change is designed to help in two cases:
1. Swipes from touchscreen into virtual key area.
2. Accidental taps in virtual key area while using on-screen keyboard.
Bug: 3089163
Change-Id: Id6733c83c2e2bc8d9553aa0e5c1fd74b741bec6e
The cause of the problem is that AudioTrack::start() can fail if it is called from a newly created
thread that has the same ID as the AudioTrack callback thread that has just been stopped and not yet exited.
This is possible as the thread ID used by the Thread class is not the TID.
The fix consists in clearing the thread ID before exiting the thread loop.
Change-Id: I66e679665c384403cb3ba2c31746f5de72d5836d
Rewrote interceptKeyBeforeQueueing to make the handling more systematic.
Behavior should be identical except:
- We never pass keys to applications when the screen is off and the keyguard
is not showing (the proximity sensor turned off the screen).
Previously we passed all non-wake keys through in this case which
caused a bug on Crespo where the screen would come back on if a soft key
was held at the time of power off because the resulting key up event
would sneak in just before the keyguard was shown. It would then be
passed through to the dispatcher which would poke user activity and
wake up the screen.
- We propagate the key flags when broadcasting media keys which
ensures that recipients can tell when the key is canceled.
- We ignore endcall or power if canceled (shouldn't happen anyways).
Changed the input dispatcher to not poke user activity for canceled
events since they are synthetic and should not wake the device.
Changed the lock screen so that it does not poke the wake lock when the
grab handle is released. This fixes a bug where the screen would come
back on immediately if the power went off while the user was holding
one of the grab handles because the sliding tab would receive an up
event after screen turned off and release the grab handles.
Bug: 3144874
Change-Id: Iebb91e10592b4ef2de4b1dd3a2e1e4254aacb697
If writeString8 is called with the following sequence:
writeString8(String8(""));
writeString8(String8("TempString"));
Then in the readString8, the 2nd String i.e. "TempString" is not read,
instead an empty string is read.
The bug comes because of the write call for String8("") where there are
no String bytes present. In the write Statement, an extra ‘\0’ is
written. During the Marshalling, Following bytes are written:
1 2 3 4 5 ...
0x0 0x0 0xB ‘T’ ‘e’ ...
The readString8 function has a check that, if String length is 0, don’t
read anything. So the first byte is read as the length for the first
string. The second byte i.e. ‘\0’ is read as the length for the second
string and hence the second string becomes empty too.
Change-Id: Id7acc0c80ae16e77be4331f1ddf69ea87e758420
* commit 'a2977c383d363e1e88a5b36230b1fa4c312807d2':
[3171580] SurfaceFlinger Bypass mode. (DO NOT MERGE)
[3171580] Add transform field to native buffers. (DO NOT MERGE)
some of these failures are not fatal and even expected in some cases
so they should not emit a dump in the log in those cases.
Change-Id: Idcfa252e3bfa9d74e27fe4ad8f8623aa01aa9c5e
Allow the overlay object to be destroyed whenever the destroy is
invoked. Currently the destroy call returns if there is an error
encountered, which results in open data and control channels.
Change-Id: I1f2ef4ebb5fb1dcabf05ab50b5bbf6e5e240a63a
This change fixes a bug in the dispatcher where the window manager
policy would incorrectly receive a key repeat count of 0 in the case
where the key repeat was generated by the hardware or driver.
Long-press on HOME was broken as a result.
Repeating keys could also get stuck down.
Bug: 3159581
Bug: 3208156
Change-Id: I1145487cfcc41a7850dba4cafc63c4a5951ace5b
This change defines a macro NUM_FRAME_BUFFERS to set
the desired number of framebuffers to be used by the UI,
instead of hard-coding 2 framebuffers.
Aditional logic has been added to handle the initialization
and destruction of NUM_FRAME_BUFFERS buffers.
Change-Id: I3a4bfec3e0f453432f2ffebf084c00f574d3be46
Signed-off-by: Rodrigo Obregon <robregon@ti.com>
Two issues:
1. First, due to an inverted conditional in the input dispatcher, we were
reporting touches as long touches and vice-versa to the power manager.
2. Power manager user activity cheek event suppression also suppresses touch
events (but not long touch or up events). As a result, if cheek event
suppression was enabled, touches would not poke the user activity timer.
However due to the above logic inversion, this actually affected long
touches. Net result, if cheek suppression was enabled in the power manager
and you held your thumb on the screen long enough, the phone would
go to sleep!
Cheek event suppression is commonly turned on when making a phone call.
Interestingly, it does not seem to get turned off afterward...
This change fixes the logic inversion and exempts touches from the cheek
suppression. The reason we do the latter is because the old behavior
was actually harmful in other ways too: a touch down would be suppressed
but not a long touch or the touch up. This would cause bizarre behavior
if you touched the screen while it was dimmed. Instead of brightening
immediately, it would brighten either when you lifted your finger or
300ms later, whichever came first.
Bug: 3154895
Change-Id: Ied9ccec6718fbe86506322ff47a4e3eb58f81834
This patch makes the dispatcher drop all of its state when it is
disabled (when the screen turns off). This ensures that the dispatcher
does not get stuck thinking a pointer is still down if the screen
turned off while the user was touching the display (such as a fat touch
while hitting the power button).
Bug: 3098344
Change-Id: If50ef5804870aa1acd3179fd4b40e3cda58dd39d
this is to allow applications to change the format of a surface's
buffer, and have it reflected in EGL; which is needed for
EGLConfig validation.
Change-Id: Iee074c30ad765881e2409c1d37450b05e561c44d
Fixed a bug where we would lose the first touch point when swiping out of
the virtual key area.
Fixed a bug where we would not send an ACTION_MOVE event in cases where
individual pointers went down/up and the remaining pointers actually moved.
This is important since many applications do not handle pointer movements
during ACTION_POINTER_DOWN or ACTION_POINTER_UP. In the case of
ACTION_POINTER_UP the movement was completely lost since all pointers were
dispatched using their old location rather than the new location.
Improved motion event validation to check for duplicate pointer ids.
Added an input source constant that was missing from the NDK api but
defined in the framework api.
Added a timestamp when reporting added/removed devices in EventHub.
Bug: 3070082
Change-Id: I3206a030f43b7616e2f48006e5a9d522c4d92e56
Compute the actual number of indices in the GAMEPAD_KEYCODES instead of
the pure size in bytes.
Bug: 3121536
Change-Id: I71edbd8bf6eff2c8cc0ea5c6845362b3d1e06466
If a ZipFileRO object is uninitialized, the hash table will not have
been initialized. This condition wasn't checked in findEntryByName.
Bug: 3121109
Change-Id: Ib696e0e7e0cb4dd0fb2e456d6a847e5e8f4fe14e
Remember, the system and main logs are
- Shared resources
- Primarily for recording problems
- To be used only for large grained events during normal operation
Bug: 3104855
Change-Id: I136fbd101917dcbc8ebc3f96f276426b48bde7b7
We now poke user activity twice: once upon dequeueing an event
for dispatch and then again just before we dispatch it. The second
poke is to compensate for the fact that it can take a few seconds to
identify the dispatch target (if the application is responding slowly)
but we want to keep the display from going to sleep for X amount of time
after the app gets a chance to actually receive the event. This mirrors
pre-Gingerbread behavior.
Removed some unnecessary code that filters user activity pokes when sending
events to KeyGuard. We don't need this because KeyGuard already tells the
power manager to disable user activity.
Bug: 3101397
Change-Id: I8c3a77601fdef8f584e84cfdd11aa79da0ff51db
This change adds a new window type for secure system overlays
created by the system itself from non-secure system overlays that
might be created by applications that have the system alert permission.
Secure views ignore the presence of secure system overlays.
Bug: 3098519
Change-Id: I8f8398f4fdeb0469e5d71124c21bedf121bd8c07
Added support for calibrating touch size for devices that report
size as an area measurement rather than as a width.
Fixed some bugs.
Bug: 3096045
Change-Id: I30a12e73752883516ed054f8af407204bca45814
