[Android] Crash in Video_Manager_DestroySurface

[Evryway]

Unity version

2022.3.62f2

Unity editor platform

Windows

AVPro Video edition

Ultra

AVPro Video version

3.3.6

Device hardware

Samsung A13, Samsung A15, Samsung A07 + 40+ known devices

Which Android OS version are you using?

13

Unity Graphics API

OpenGLES 3

Video API

MediaPlayer

Texture format

BGRA

Audio output

System Direct

Any other Media Player component configuration required to reproduce the issue.

No response

Which output component(s) are you using?

Display uGUI

Any other component configuration required to reproduce the issue.

No response

The issue

The latest release of our application using AVPro 3.3.6 on Android appears to have new crashes related to DestroySurface. Android Vitals on play store suggests the app is foregrounded when this occurs. There are no explicit calls to CloseVideo/CloseMedia other than during app shutdown flow, which I don't think has happened.

Play store suggests it's happening on a wide variety of devices - I can provide a detailed list if that's useful.

I'm unable to replicate this crash on the various devices I have to hand to test on so unfortunately the best I have is the play store trace:

arm64/libAVProVideo2Native.so (Java_com_renderheads_AVPro_Video_Manager_DestroySurface+68)
data/misc/apexdata/com.android.art/dalvik-cache/arm64/boot.oat (art_jni_trampoline+108)
arm64/base.odex (com.renderheads.AVPro.Video.Player_ExoPlayer.CloseVideoOnPlayer+332)
arm64/base.odex (com.renderheads.AVPro.Video.Player_Base._CloseVideo+948)
apex/com.android.art/lib64/libart.so (nterp_helper+4016)
base.apk (com.renderheads.AVPro.Video.Player_Base._Deinitialise+16)
apex/com.android.art/lib64/libart.so (nterp_helper+3924)
base.apk (com.renderheads.AVPro.Video.Player_Base.access$000)
apex/com.android.art/lib64/libart.so (nterp_helper+52)
base.apk (com.renderheads.AVPro.Video.Player_Base$1.run+4)
data/misc/apexdata/com.android.art/dalvik-cache/arm64/boot.oat (android.os.Handler.dispatchMessage+68)
data/misc/apexdata/com.android.art/dalvik-cache/arm64/boot.oat (android.os.Looper.loopOnce+968)
data/misc/apexdata/com.android.art/dalvik-cache/arm64/boot.oat (android.os.Looper.loop+824)

Any ideas how I can rectify this please?

Thanks!

Media information

No response

Logcat output

[MorrisRH]

Without reproduction steps we're unlikely to be able to figure out what's going on.
Going by the call stack the player instance is in the process of being destroyed so if you can provide any further information about when and how this is occurring would be very helpful.

The full list of devices might be handy, we don't have access to any of the devices you listed but do have a couple of Samsung A class devices to test against so will start there.

[Evryway]

Thanks for the response @MorrisRH. I fully appreciate without repro steps this is going to be tricky to find,.

I'm digging through the code currently to try and figure out the flow.
There's two UI screens that share a single MediaPlayer instance. the first screen has one DisplayUGUIs, the second screen has two. These screens are always resident in memory until the app closes.
Switching between the screens at runtime will start video playback (one specific video per screen, local to the device). The call to start the video is _mediaPlayer.OpenMedia(...).
Going between the screens one way (A->B) does no specific shutdown flow. Going the other way is calling _mediaPlayer.Stop() followed by _mediaPlayer.CloseMedia() on video B, then OpenMedia() on video A. CloseMedia happens on frame X, then OpenMedia happens on frame X+1.

Are there any specific timing issues or guards I should be aware of / putting in place to make this flow safe? should I always be calling CloseMedia prior to a new OpenMedia?

would using multiple MediaPlayer instances be a better pattern to guard against this issue?

List of known devices (non-exhaustive, the live build has been rolled back while we investigate the issue)
Most of these look pretty low spec, but not all.

Device Name | Chipset
Aonn 100135794 | Mediatek MT8786
ATOZEE CP10_Pro | Rockchip RK3562C
Blackview Tab_A6_Kids | Allwinner A523
google shiba | Google Tensor G3
HONOR HNCRT-M2 | Mediatek MT6883
HONOR HNGFY-M | Mediatek MT6765V/XBA
HONOR HNNDL-Q | QTI SM6225
HONOR HNVNE-M | Mediatek MT6762
HONOR HNWDY-M | Mediatek MT6765V/XBA
Infinix Infinix-X6882 | Mediatek MT6789
Jeazans KT1016 | Allwinner A523
Lenovo TB301FU | Mediatek MT8768A
motorola aito | Mediatek MT6878
motorola kansas | Mediatek MT6835
motorola lamu | Mediatek MT6769V/CB
motorola manila | Spreadtrum T760
Nothing Spacewar | Qualcomm SM7325
OnePlus OP5D55L1 | QTI SM8750
onn 100135920 | Mediatek MT8786
OPPO OP4F4DL1 | Mediatek MT6853V/TNZ
OPPO OP574FL1 | Mediatek MT6769
OPPO OP5B16L1 | Qualcomm SM6115
POCO gust | Mediatek MT6769Z
realme RE5C7CL1 | Mediatek MT6789V/CD
Redmi flare | Mediatek MT8786V/CA
Redmi spark | Mediatek MT8786V/CA
Redmi spes | QTI SM6225
Redmi xun | QTI SM6225
samsung a05m | Mediatek MT6769V/WY
samsung a07 | Mediatek MT6789V/CD
samsung a13 | Samsung Exynos 850
samsung a14xm | Mediatek MT6833V/NZA
samsung a15 | Mediatek MT6789V/CD
samsung a15x | Mediatek MT6835V/ZA
samsung a34x | Mediatek MT6877V/TTZA
samsung dm3q | QTI SM8550
samsung e3q | QTI SM8650
samsung gta4xlvewifi | Qualcomm SM7125
samsung gta7litewifi | Mediatek MT8768N
samsung gta9p | QTI SM6375
samsung q4q | QTI SM8475
samsung r11s | Samsung s5e9925
TCL Beryl_VZW | Mediatek MT6835
vivo V2250 | Qualcomm SM7325
Vortex C24L | Spreadtrum SC9863A
weelikeit C19W | Allwinner A523
WOWDEFU C18W_EEA | Allwinner A523

[Chris-RH]

What is the difference between the previous version that works and the new one that doesn't?
What is the difference between the videos that it is switching between? (resolution/codec/fps/bitrate etc)

[Evryway]

The previous version that worked only had one UI screen that was showing videos. The new version flips between two UI screens.
The previous version was on 3.3.4, the new version is on 3.3.6.
video A is 1280x720 25fps 1300kbps H264, no audio.
video B is 1920x1080 25fps 3000kbps H265 with audio.

I'm assuming this is the cause (as I don't have a local repro case). Could you give any more details on the likely code paths that would cause _DestroySurface please? If there's somewhere else I should be looking, that would be great info.