[Android]: API 'mesh.path:update' index data not update

[zhiyangyou]

• Device: Android Simulator On Windows (夜神模拟器 6.6.1.2)
• OS: Android5 5.1.1
• solar2D code Version: branch : master commit : e5217ec

Describe the bug
on windows simulator ， API 'mesh.path:update' work normally
on android platform , API 'mesh.path:update' does not work, seem to that IBO has not been updated
this bug will show by gif image

test demo
assets2.zip

bug

bug fixed??

Since I am not familiar with the source code, I am not sure if this modification is correct,
in Rtt_GlGeometry.cpp line:301,
bind target IBO
bufferSubData IBO
bind GL_ELEMENT_ARRAY_BUFFER to 0

Looking forward to your reply , thanks

[Shchvova]

Hey. May I ask to try in latest build?

[zhiyangyou]

Hey. May I ask to try in latest build?

Thank you for your reply
Bug still exist

[zhiyangyou]

In my application, I intergrate a spine animation system to engine .
In order to update mesh data more quickly, I take part of the business logic to c + + layer, direct access ShapeObject.ClosePath.TesselatorMesh and update it's mesh data

[depilz]

Hi @zhiyangyou , you are right, I already fixed this in Spine but haven't shared it since it is not the fix we really need.

Which Spine Runtime version are you using?

[zhiyangyou]

Thank you for reply @depilz

The "spine" version I use is:
spine-runtimes
Branch: 4.0
commit:b8da2fa27bd7fbdc09c041c12424a2d8df70e041
I integrated the spine-cpp library into my application and re-implemented the adapter between spine-runtime.lib and corona engine using c++, like spine-solar2d/spine.lua

Because ShapeObject was reused and index data was not pushed to the GPU side during the Verts UVs Indices update.
I am now looking forward to the author's repair

And I suggest that spine with cliperAttachment be added to the "solar2d-spine" test case
Such as "coin"
I used solar2d-spine of the above version for this picture

[depilz]

Woow! For real? Having Spine running in C++ would be an incredible boost in performance!
Did you made it as a plugin?

I'm working with Spine 3.9, but I'm going to update 4.0 too and share the fix in a moment.

[depilz]

@zhiyangyou This is it:

spine-4-0.txt

[zhiyangyou]

@zhiyangyou This is it:

spine-4-0.txt

Thanks for sharing
I see the areEqual method, which avoids index updates

[zhiyangyou]

Woow! For real? Having Spine running in C++ would be an incredible boost in performance! Did you made it as a plugin?

I'm working with Spine 3.9, but I'm going to update 4.0 too and share the fix in a moment.

I modified the Rtt_LuaLibDisplay.cpp file and added the display.newspine() method,
There are no plugin yet, also, I don't think it's easy to plug , because of the source code modification

Using spine-cpp did improve performance a lot in my tests

1. Memory usage is lower
2. Avoid a lot of Lua gc and memory allocation time
3. Avoid the time consumption of data transfer in Lua-C interactive stack
4. Use ShapeObject pool to avoid the time spent constructing many objects (paint Mesh) , but pool cost more memory usage , special in spine with clipper

[depilz]

That's exactly what I wanted to do! But didn't know C++ enough to do that task alone.

When I made the tests, most of the delay was caused by running the runtime and only a small portion by the drawing itself, so migrating to C++ was my hope.

If you have a public fork of this I would love to collaborate and help making this officially part of the engine.

[zhiyangyou]

That's exactly what I wanted to do! But didn't know C++ enough to do that task alone.

When I made the tests, most of the delay was caused by running the runtime and only a small portion by the drawing itself, so migrating to C++ was my hope.

If you have a public fork of this I would love to collaborate and help making this officially part of the engine.

Currently in development,there are still some issues that have not been dealt with , using Corona's memory management, disk IO, assetsPath , bug fix , .etc
It may take some times, then I will push the code to the repository I forked.And I will notify you here. Very little code that works

[ggcrunchy]

Hi. @depilz brought this up on Discord and there were some comments there: https://discord.com/channels/721785436195782677/721785436195782680/973953535035199498

I'm StarCrunch there and was wondering some things myself. Do you know all the parts where you need to integrate with the engine, versus where it's convenient?

[ggcrunchy]

@zhiyangyou Hi. Something I saw just now:

Vertex array objects are disabled on Android, so you go through this path: https://github.com/coronalabs/corona/blob/master/librtt/Renderer/Rtt_GLGeometry.cpp#L282

But that's missing the index update bits from further up: https://github.com/coronalabs/corona/blob/master/librtt/Renderer/Rtt_GLGeometry.cpp#L268 This seems to be exactly your situation.

[zhiyangyou]

Hi. @depilz brought this up on Discord and there were some comments there: https://discord.com/channels/721785436195782677/721785436195782680/973953535035199498

I'm StarCrunch there and was wondering some things myself. Do you know all the parts where you need to integrate with the engine, versus where it's convenient?

hi @depilz corona_spine_cpp_adapter demo code https://github.com/zhiyangyou/corona/blob/master_spine_cpp_intergration_test/Corona%26SpineCppREADME.md
Frame rate comparison
https://github.com/zhiyangyou/corona/tree/master_spine_cpp_intergration_test/doc_images

I can not access this discord channel ， discord user name zhiyangyou#5364

[zhiyangyou]

My problem is solved, thank you for your kind help

[depilz]

Wooah!! That's amazing! Thanks for this big contribution @zhiyangyou.

The discussion was about what is the best way to make this officially part of the engine, and that's probably making it a plugin, and I'm gonna work on it.

[Shchvova]

We got set of pull requests to make it possible for this to be a plug-in.
Probably should get back to them.