Develop Reference

C++11 std::chrono::steady_clock issue on Android

I have been using std::chrono::steady_clock for interval calculation in an application i am making for Android platform.
Code:
// On application start
auto timeSinceEpoch = std::chrono::steady_clock::now().time_since_epoch();
auto timeInSec = std::chrono::duration_cast<seconds>(timeSinceEpoch).count();
log("On Enter Start Time Point - %lld", timeInSec);
Output:
On Enter Start Time Point - 521
Now i switch off the phone and restart the phone. I run my application and this time Output is:
On Enter Start Time Point - 114
As per definition at cppreference.com
"Class std::chrono::steady_clock represents a monotonic clock. The time points of this clock cannot decrease as physical time moves forward."
How is the output when i restart the phone giving lesser value?
If anyone has faced this issue please help me out here. Thanks!!

The formal requirement for a steady clock is that the result of a call to now() that happens before another call to now() is always less than or equal to the result of the second call. The happens before relationship only applies to actions within a program run. A steady clock is not required to be steady across different invocations of a program.

On Android, AFAICT steady_clock is the same as (from Java) System.Clock.elapsedRealtime, which resets to zero on boot -- https://developer.android.com/reference/android/os/SystemClock.html
I'm totally failing to dig up the source code for clock_gettime, though. https://android.googlesource.com/platform/ndk.git/+/43255f3d58b03cd931d29d1ee4e5144e86e875ce/sources/cxx-stl/llvm-libc++/libcxx/src/chrono.cpp#124 shows it calling clock_gettime(CLOCK_MONOTONIC), but I'm not sure how to penetrate the veil from there.

how to deal with this case: andorid epoll_wait return -1 and errno=4 used ndk

I am writing network communication program with Android ndk, using epoll.
I found the method ‘epoll_wait’ woken not very accurate
while(1){
struct epoll_event events[3];
log_string("epoll_wait start");//here will print start time
events_len = epoll_wait(_epoll_fd, events, 3, 20 * 1000);// wait 20 second,for test,I use pipe instead of socket,monitor a pipe EPOLLIN event
if (events_len <= 0) {
log_string("epoll_wait end events_len=%d,errno=%d", events_len, errno);//Normally,the events_len always is 0,and errno is 0
}
}
The above code runs on the PC(like Ubuntun PC) is very normal,as expected.
If it runs on Android Phone（use Android Service , separate thread to run） is as expected at first.
After some time,epoll_wait becomes not very accurate,sometimes got -1 and errno=4,sometimes waited very long.
So I only know that phenomenon, but do not know why.
Can you tell why and tell me the best practices for use android epoll?
thx

4 is EINTR, which means your app got a signal. This isn't really an error, just restart epoll.
Regarding "waited very long", does your app hold at least a partial wakelock?

Gideros image dynamic load and delete from memory

I would like to make an mobile application, what contains a lot of picture
My question how can I dynamically open the picture and delete from memory?
I tested this:
a = Texture.new("a.jpg")
print(Application:getTextureMemoryUsage()) -- write x
a = nil
print(Application:getTextureMemoryUsage()) -- write x again
Thanks for help.

Problem is that garbage is not collected right away and that is why memory is not freed right away.
You could try calling collectgarbage() couple of times to force it as:
print(math.floor(collectgarbage("count")))
collectgarbage()
collectgarbage()
collectgarbage()
print(math.floor(collectgarbage("count")))

ALSA - unmuting devices?

I have been trying to capture audio, within a native linux program running on an Android device via adb shell.
Since I seemed to be getting only (very quiet) noise, i.e. no actual signal (interestingly, an Android/Java program doing similar did show there was a signal on that input),
I executed alsa_amixer, which had one entry that looked like the right one:
Simple mixer control 'Capture',0
Capabilities: cvolume cswitch penum
Capture channels: Front Left - Front Right
Limits: Capture 0 - 63
Front Left: Capture 31 [49%] [0.00dB] [off]
Front Right: Capture 31 [49%] [0.00dB] [off]
"off". That would explain the noise.
So I looked for examples of how to use alsa_amixer to unmute the channels, I found different suggestions for parameters like "49% on" or "49% unmute", or just "unmute" none of which works. (if the volume% is left out, it says "Invalid command!", otherwise, the volume is set, but the on/unmute is ignored)
I also searched how to do this programatically (which I'll ultimately need to do, although the manual approach would be helpful for now), but wasn't too lucky there.
The only ALSA lib command I found which sounds like it could do something like that was "snd_mixer_selem_set_capture_switch_all", but the docs don't day what the parameter does (1/0 is not on/off, I tried that ;) )

The manual approach to set these things via alsa_amixer does work - but only if android is built with the 'BoardConfigCommon.mk' modified, at the entry: BOARD_USES_ALSA_AUDIO := false, instead of true.
Yeah, this will probably disable ALSA for android, which is why it wouldn't meddle with the mixer settings anymore.
To you android programmers out there, note that this is a very niche use case of course, as was to be expected by my original post to begin with.
This is not what most people would want to do.
I just happen to tinker with an android device here in unusual ways ;-)

Just posting the code as question giver suggested, also don't like external links.
#include <alsa/asoundlib.h>
int main()
{
snd_mixer_t *handle;
snd_mixer_selem_id_t *sid;
snd_mixer_open(&handle, 0);
snd_mixer_attach(handle, "default");
snd_mixer_selem_register(handle, NULL, NULL);
snd_mixer_load(handle);
snd_mixer_selem_id_alloca(&sid);
snd_mixer_selem_id_set_index(sid, 0);
snd_mixer_selem_id_set_name(sid, "Capture");
snd_mixer_elem_t* elem = snd_mixer_find_selem(handle, sid);
snd_mixer_selem_set_capture_switch_all(elem, 0);
snd_mixer_selem_set_capture_dB_all(elem, 0, 0);
snd_mixer_close(handle);
}

Memory metrics in android apps

I wrote Android application and it have some strange problems with Out Of Memory exception, which appears randomly. Yes, I know that problems with OOM exception usually because of images, and I used all what I can to avoid this problem.
The only way I could think to find the place where spend the memory is putting the logs with information about memory everywhere. But I really confused about which values ​​do I need.
I used next values:
Runtime.getRuntime().maxMemory()
Runtime.getRuntime().freeMemory()
Runtime.getRuntime().totalMemory()
Debug.getNativeHeapAllocatedSize()
Debug.getNativeHeapFreeSize()
Debug.getNativeHeapSize()
And before OOM exception I have next values:
maxMemory: 57344K
freeMemory: 9581K
totalMemory: 22407K
NativeHeapAllocatedSize: 34853K
NativeHeapFreeSize: 302K
NativeHeapSize: 40060K
lowMemory false
In this question Android Bitmap Limit - Preventing java.lang.OutOfMemory I see that used compeering ((reqsize + Debug.getNativeHeapAllocatedSize() + heapPad) >= Runtime.getRuntime().maxMemory()) ant in this blog some strange information:
Debug.getNativeHeapFreeSize(); The free size is the amount of memory
from the heap that is not being used, because of fragmentation or
provision.
Also I can not understand haw can be OOM exception if (totalMemory: 22407K) much less than (maxMemory: 57344K).
Please help me understand how use this values.