Making a JNI call from tick tasks is not permitted as the underlying thread
is not attached to the Java VM. This is an error and crashes in the emulator
(which has stricter JNI checks enabled by default).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30173 a1c6a512-1295-4272-9138-f99709370657
* Cleanup RockboxService.java by moving the battery and
headphone monitors to separate classes and detaching their instances
* Move those monitors and RockboxTelephony.java into a new
monitors subdirectory
* Call those monitors all the same from native code by creating
the objects there
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30160 a1c6a512-1295-4272-9138-f99709370657
Listen to headphone plug events. There are currently two glitches with this:
- Android takes a while until it reports the unplug event, so there will be
some delay until playback gets paused. This is an Android limitation.
- Rockbox debounces headphone state changes for one second. Therefore playback
will shortly be routed to the speaker on unplug until Rockbox does the actual
pause.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29956 a1c6a512-1295-4272-9138-f99709370657
The environment is fine to share in general, just not across OS threads, so it's only needed
for functions which are possibly called from multiple OS threads (only 1 currently).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29601 a1c6a512-1295-4272-9138-f99709370657
The implementation wasn't really safe (sem_post from signal handler), and the new one simplifies things a bit.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29575 a1c6a512-1295-4272-9138-f99709370657
The added complexity wasn't needed for most subsystems, as main() never returns
so local references can't be freed.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29570 a1c6a512-1295-4272-9138-f99709370657
I tried to move the #ifdefs and the code
in firmware/powermgmt.c around and it was still
a big mess for hosted applications (RaaA/sim builds).
Create our own "power" thread as recently discussed on IRC.
Fixes the sleep timer for RaaA.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29501 a1c6a512-1295-4272-9138-f99709370657
* Re-create RockboxFramebuffer instance with every time there's a new Activity.
* Also, allow Rockbox to be started via multimedia buttons, immediately starting playback if wanted.
We don't need to keep the fb instance around when it backround, and it makes us less depending on it and the activity (less race conditions). And this is how you usually do it in Android apps.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29384 a1c6a512-1295-4272-9138-f99709370657
That enables drawing from outside the apps UI thread, i.e. from within the Rockbox native thread, without needing synchronization means, and adds determinism as to when the draw happens.
It simplifies the drawing routines and adds a convinient way of detecting whether drawing should happen or not (surfaceCreated/Destroyed).
It also restores max. fps on my phone which went down drastically with the gingerbread(CM7) update.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29333 a1c6a512-1295-4272-9138-f99709370657
This replaces SDL threads with real cooperative threads, which are less cpu intensive and allow priority scheduling.
The backend for context switching is dependant on the host (sigaltstack/longjmp on Unix, Fibers on Windows).
configure has options to force or disallow SDL threads.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29327 a1c6a512-1295-4272-9138-f99709370657
First, it add the ability to tagcache to walk through multiple search roots.
Second, it adds symlinks targets to the search roots if they're are not inside any of the current search roots, otherwise the symlink is ignored (unless it's a file).
The default search root is still /, so no search root will be actually added.
But the tagcache now isn't trapped by recursive symlinks anymore and successfully builds, and it's prepared for a future music directory setting.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28884 a1c6a512-1295-4272-9138-f99709370657
The scrolling engine could lock out the main thread which may be waiting for its update to finish (wakeups are not reentrant).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28821 a1c6a512-1295-4272-9138-f99709370657
test_fps can also now report reasonable numbers: ~62 fps for both 1/1 and 1/4 updates (was 300-400 previously).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28728 a1c6a512-1295-4272-9138-f99709370657
* For some reason, the polling methid is much more inefficient than I thought. According to htop it caused up to 15% CPU load on some phones (e.g. Galaxy S).
The event based causes no CPU load.
Rockbox' idle CPU load is now back to 0%, while it was previously dominated by polling the call state.
* Also stop on outgoing calls (no need to explicitely pause for making a call anymore).
* Factor out the detection mechanism to separate files.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28564 a1c6a512-1295-4272-9138-f99709370657
Remove the hardcoded (and way too small) scroll threshold (the distance moved in pixels before we think the users wants to scroll) and replace it with something based on the actual DPI of the screen.
On Android we call the API for that, on other touchscreens we reimplemented Android's formula (as of 2.2) and calculate it.
Flyspray: 11727
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28548 a1c6a512-1295-4272-9138-f99709370657
Also make the back button go back to the OS home from the rockbox main menu
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28475 a1c6a512-1295-4272-9138-f99709370657
Remove some @Override annotations to make the Java code build with certain javac versions.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28386 a1c6a512-1295-4272-9138-f99709370657
Exclude compilation of firmware/debug.c for hosted platforms.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28125 a1c6a512-1295-4272-9138-f99709370657
