Fatal signal 6 when calling AbgrBuffer.allocate()

[Liviu Timar]

Hello,

I am getting a crash with the following stacktrace. From what I can tell, it starts with a call to: AbgrBuffer.allocate(width, height) in my code.

JNI DETECTED ERROR IN APPLICATION: non-zero capacity for nullptr pointer: 3686400
in call to NewDirectByteBuffer
from java.nio.ByteBuffer io.github.crow_misia.libyuv.Yuv.allocNativeBuffer(int)
at AbgrBuffer.allocate(width, height)
at io.github.crow_misia.libyuv.Yuv.allocNativeBuffer(Native method)
at io.github.crow_misia.libyuv.BufferExtKt.createByteBuffer(BufferExt.kt:21)
at io.github.crow_misia.libyuv.AbgrBuffer$Factory.allocate(AbgrBuffer.kt:110)

Fatal signal 6 (SIGABRT), code -1 (SI_QUEUE) in tid 3663 (CaptureThread), pid 2731 (o.android.debug)
~ Channel is unrecoverably broken and will be disposed!

I am doing this call for each frame of a video. The crash appears after 10 minutes of use.

Do you have any idea about what the problem could be? Let me know if you need more details.

Thank you,
Liviu

[Zenichi Amano]

Seems to be a memory leak.
Instances should be reused instead of creating one for each frame.

[Liviu Timar]

Thanks for answering!

How can I reuse the buffer, taking into account that the width and height might change from a frame to another?

[Zenichi Amano]

Maybe an implementation similar to this.

private var buffer: AbgrBuffer

fun onFrame(image: Image) {
  if (image.width != buffer.width || image.height != buffer.height) {
    buffer.close()
    buffer = AbgrBuffer.allocate(image.width, image.height)
  }

  // image processing.
}

[Zenichi Amano]

please reopen if there is more information.