Question: shouldn't all instances of Pointer be deallocated to release native memory?

[netbrain]

I see that this is not the case for this line, is this causing a memory leak?

javacv/src/main/java/org/bytedeco/javacv/OpenCVFrameConverter.java

Line 93 in da3bb9c

&& new Pointer(frame.image[0].position(0)).address() == img.imageData().address()

[netbrain]

Also i can't seem to find any release of memory for these two lines:

javacv/src/main/java/org/bytedeco/javacv/FFmpegFrameRecorder.java

Lines 167 to 168 in 6cc1da8

	this.video_pkt = new AVPacket();
	this.audio_pkt = new AVPacket();

Im guessing the releaseUnsafe method should call av_free_packet()?

[saudet]

The first one doesn't cause allocation of any native memory, so that's alright.

As for video_pkt and audio_pkt, the default allocator registers a deallocator for the garbage collector to call at some point after they become unreachable. We could deallocate them explicitly on release, but they won't cause a memory leak.

Are you having any issues with a memory leak somewhere?

[netbrain]

Well, im not sure. I think maybe I have fixed it, but needs further testing :) Any hints for how to track down memory leaks in JNI? so far I have tried BCC's memleak utility. but all in all, allocations looks fairly good. I've also tried using AspectJ and instrumenting JavaCPP's Pointer class and spying on set on the address field and using that information to create a summary of allocs & deallocs, not the greatest set of results though.. Haven't tried valgrind yet, but i guess that will be next if i still see crashes due to memory.

[saudet]

If we set, for example, the "org.bytedeco.javacpp.logger.debug" system property to "true", it will display allocations and deallocations on the console, but yes, we need some other tool to analyze that log. If you create some Java code for that, please send a pull request to have it integrated into JavaCPP itself! Thanks

[saudet]

BTW, I think I've fixed all the memory leaks occurring in FFmpegFrameGrabber and FFmpegFrameRecorder.
Please give it a try with the snapshots: http://bytedeco.org/builds/

[saudet]

PointerBufferPoolMXBean from @matteodg released with JavaCPP 1.5.4 should make it easier to debug.
If you still have issues with this though, please let me know!