mediapipe as dependency using bazel

[davidakr]

Hi,
I am building a C++ Application on Ubuntu which should use mediapipe.
How do I configure the mediapipe framework as a remote dependency in a bazel project?

[mgyong]

@davidakr Pls check out the helloworld example in the mediapipe_addon repo of how to create C++ app with MediaPipe as remote dependency in bazel

[davidakr]

The hello_world example works just fine.
I am trying to build the hand tracking with GPU support enabled but I cannot build the @mediapipe//mediapipe/graphs/hand_tracking:mobile_calculators. In my application I want to use this feature. I run into the following issue:

In file included from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.h:21:0,
from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:16:
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc: In function 'int tflite::gpu::gl::{anonymous}::GetConfig(EGLDisplay, const EGLint*, void**)':
external/org_tensorflow/tensorflow/lite/delegates/gpu/common/status.h:68:18: error: request for member 'ok' in 'status2', which is of non-class type 'const int'
if (!status2.ok()) return status2;
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:33:3: note: in expansion of macro 'RETURN_IF_ERROR'
RETURN_IF_ERROR(GetOpenGlErrors());
^~~~~~~~~~~~~~~
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:35:69: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return InternalError("No EGL error, but eglChooseConfig failed.");
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:37:19: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return OkStatus();
^
In file included from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.h:21:0,
from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:16:
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc: In function 'int tflite::gpu::gl::{anonymous}::CreateContext(EGLDisplay, EGLContext, EGLConfig, tflite::gpu::gl::EglContext*)':
external/org_tensorflow/tensorflow/lite/delegates/gpu/common/status.h:68:18: error: request for member 'ok' in 'status2', which is of non-class type 'const int'
if (!status2.ok()) return status2;
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:50:3: note: in expansion of macro 'RETURN_IF_ERROR'
RETURN_IF_ERROR(GetOpenGlErrors());
^~~~~~~~~~~~~~~
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:52:70: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return InternalError("No EGL error, but eglCreateContext failed.");
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:55:19: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return OkStatus();
^
In file included from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.h:21:0,
from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:16:
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc: In member function 'int tflite::gpu::gl::EglContext::MakeCurrent(EGLSurface, EGLSurface)':
external/org_tensorflow/tensorflow/lite/delegates/gpu/common/status.h:68:18: error: request for member 'ok' in 'status2', which is of non-class type 'const int'
if (!status2.ok()) return status2;
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:98:3: note: in expansion of macro 'RETURN_IF_ERROR'
RETURN_IF_ERROR(GetOpenGlErrors());
^~~~~~~~~~~~~~~
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:100:68: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return InternalError("No EGL error, but eglMakeCurrent failed.");
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:102:19: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return OkStatus();
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc: In function 'int tflite::gpu::gl::CreateConfiglessContext(EGLDisplay, EGLContext, tflite::gpu::gl::EglContext*)':
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:112:70: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return UnavailableError("EGL_KHR_no_config_context not supported");
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc: In function 'int tflite::gpu::gl::CreateSurfacelessContext(EGLDisplay, EGLContext, tflite::gpu::gl::EglContext*)':
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:120:67: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return UnavailableError("EGL_KHR_create_context not supported");
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:123:72: error: cannot convert 'tflite::gpu::Status' to 'int' in return
return UnavailableError("EGL_KHR_surfaceless_context not supported");
^
In file included from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.h:21:0,
from external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:16:
external/org_tensorflow/tensorflow/lite/delegates/gpu/common/status.h:68:18: error: request for member 'ok' in 'status2', which is of non-class type 'const int'
if (!status2.ok()) return status2;
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:128:3: note: in expansion of macro 'RETURN_IF_ERROR'
RETURN_IF_ERROR(GetConfig(display, attributes, &config));
^~~~~~~~~~~~~~~
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc: In function 'int tflite::gpu::gl::CreatePBufferContext(EGLDisplay, EGLContext, tflite::gpu::gl::EglContext*)':
external/org_tensorflow/tensorflow/lite/delegates/gpu/common/status.h:68:18: error: request for member 'ok' in 'status2', which is of non-class type 'const int'
if (!status2.ok()) return status2;
^
external/org_tensorflow/tensorflow/lite/delegates/gpu/gl/egl_context.cc:146:3: note: in expansion of macro 'RETURN_IF_ERROR'
RETURN_IF_ERROR(GetConfig(display, attributes, &config));

I was trying to build it with several different tensorflow versions but no success.
Building the provided examples works just fine
Any idea how to fix this?

[jiuqiant]

Regrading your issue, this discussion is somehow related: #305 (comment)

[davidakr]

this solved my issue.