Track external android format

[ziga-lunarg]

The external format from android must be tracked, because in GetImageResourceSizesOptimal we create temporary images from the same format to calculate the resource size

If format is VK_FORMAT_UNDEFINED and the external android format is not provided it will crash on android

[ci-tester-lunarg]

CI gfxreconstruct build queued with queue ID 245923.

[ci-tester-lunarg]

CI gfxreconstruct build queued with queue ID 245925.

[ci-tester-lunarg]

CI gfxreconstruct build # 4725 running.

[bradgrantham-lunarg]

Looks good to me except possibly using the convenience function if it fits.

What would a minimal test program for this case look like?

[ci-tester-lunarg]

CI gfxreconstruct build # 4725 passed.

[ci-tester-lunarg]

CI gfxreconstruct build queued with queue ID 246259.

[ziga-lunarg]

Looks good to me except possibly using the convenience function if it fits.

What would a minimal test program for this case look like?

A minimal test program would have to create an AHardwareBuffer and then create an image with an external android format, and allocate device memory with the AHardwareBuffer and bind it to the image. And I think this only happens in a trimmed capture. This should reproduce it, but not 100% sure.

[ci-tester-lunarg]

CI gfxreconstruct build # 4729 running.

[ci-tester-lunarg]

CI gfxreconstruct build # 4729 passed.

[bradgrantham-lunarg]

Looks good to me except possibly using the convenience function if it fits.
What would a minimal test program for this case look like?

A minimal test program would have to create an AHardwareBuffer and then create an image with an external android format, and allocate device memory with the AHardwareBuffer and bind it to the image. And I think this only happens in a trimmed capture. This should reproduce it, but not 100% sure.

Do we have a capture we could put in CI that replicates this or could you quickly write a program to replicate this?

[ziga-lunarg]

Looks good to me except possibly using the convenience function if it fits.
What would a minimal test program for this case look like?

A minimal test program would have to create an AHardwareBuffer and then create an image with an external android format, and allocate device memory with the AHardwareBuffer and bind it to the image. And I think this only happens in a trimmed capture. This should reproduce it, but not 100% sure.

Do we have a capture we could put in CI that replicates this or could you quickly write a program to replicate this?

I do have a capture, but I've realized this fix is not enough. There is another problem later in the capture, that I thought was unrelated, but it seems that it is.