# Help getting CMake to install a library created by custom command

**URL:** https://discourse.cmake.org/t/help-getting-cmake-to-install-a-library-created-by-custom-command/4898
**Category:** Code
**Created:** [January 24, 2022, 8:18pm UTC](https://discourse.cmake.org/t/help-getting-cmake-to-install-a-library-created-by-custom-command/4898 "2022-01-24T20:18:34Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![mathomp4](https://discourse.cmake.org/user_avatar/discourse.cmake.org/mathomp4/32/176_2.png) [@mathomp4](https://discourse.cmake.org/u/mathomp4)
#### Post date: [January 24, 2022, 8:18pm UTC](https://discourse.cmake.org/t/help-getting-cmake-to-install-a-library-created-by-custom-command/4898/1 "2022-01-24T20:18:34Z")

</div>

I have an odd situation that I’m struggling with the CMake for. I work on a big Fortran source code that now has a need to build a shared object library and include file where they are created by _Python_ and not by Fortran. We have a “main” library (made by Fortran) here which works just fine, called `libFVdycoreCubed_GridComp.a` (or .so, I suppose, but static for now). It’s this “extra” library that’s getting me.

Now, I can build the library at the moment with some horribly ugly CMake that creates an `INTERFACE` library (not sure if that’s what I should do?) and all seems to build. The library is made, all the executables that need it are satisifed, etc.

Now, at the end of my `install` step, I see the library in the `build` directory (aka `CMAKE_CURRENT_BINARY_DIR` for this file):

```auto
$ ls -l src/Components/@GEOSgcm_GridComp/GEOSagcm_GridComp/GEOSsuperdyn_GridComp/@FVdycoreCubed_GridComp/
total 610
drwxr-xr-x 4 mathomp4 g0620 512 Jan 24 13:53 @fvdycore/
drwxr-xr-x 10 mathomp4 g0620 8192 Jan 24 13:53 CMakeFiles/
-rw-r--r-- 1 mathomp4 g0620 581 Jan 24 12:26 CTestTestfile.cmake
-rw-r--r-- 1 mathomp4 g0620 70632 Jan 24 13:53 Makefile
-rw-r--r-- 1 mathomp4 g0620 13425 Jan 24 13:53 cmake_install.cmake
-rw-r--r-- 1 mathomp4 g0620 72548 Jan 24 13:54 fv_dynamics_interface_py.c
-rw-r--r-- 1 mathomp4 g0620 2236 Jan 24 13:53 fv_dynamics_interface_py.h
-rw-r--r-- 1 mathomp4 g0620 60104 Jan 24 13:54 fv_dynamics_interface_py.o
-rwxr-xr-x 1 mathomp4 g0620 54096 Jan 24 13:54 libfv_dynamics_interface_py.so*
drwxr-xr-x 3 mathomp4 g0620 512 Jan 24 13:53 scripts/

```

But I have one last thing I can’t figure out. I want/need this `libfv_dynamics_interface_py.so` file to end up in `install/lib` and I can’t seem to get it there. For example, I’ve tried:

```cmake
install(TARGETS fv_dynamics_interface_py
  EXPORT ${PROJECT_NAME}-targets
  LIBRARY DESTINATION ${CMAKE_INSTALL_PREFIX}/lib
  )

```

but I’m guessing `install` doesn’t really see it as an “actual” library. Okay, I also tried adding an explicit:

```cmake
install(FILES
  ${CMAKE_CURRENT_BINARY_DIR}/libfv_dynamics_interface_py.so
  DESTINATION ${CMAKE_INSTALL_PREFIX}/lib
)

```

But, nope, not there at `make install`.

Maybe this is all due to the fact I sort of “hijacked” the `add_library(INTERFACE)` to try and get this to work?

I’m sure I’m missing something obvious…but it’s evading me!

---

<div class="post-metadata">

### Author: ![mathomp4](https://discourse.cmake.org/user_avatar/discourse.cmake.org/mathomp4/32/176_2.png) [@mathomp4](https://discourse.cmake.org/u/mathomp4)
#### Post date: [January 24, 2022, 9:21pm UTC](https://discourse.cmake.org/t/help-getting-cmake-to-install-a-library-created-by-custom-command/4898/2 "2022-01-24T21:21:26Z")

</div>

Okay. I have…something. I tried adding a `add_custom_command(POST_BUILD)` on my library, but I got a "`POST_BUILD` not allowed on `INTERFACE` library" error. But, my library depended on a custom\_target that made the library and:

```cmake
set(FVDYN_INTERFACE_LIBRARY
  ${CMAKE_CURRENT_BINARY_DIR}/libfv_dynamics_interface_py.so
  )

set(FVDYN_INTERFACE_HEADER_FILE
  ${CMAKE_CURRENT_BINARY_DIR}/fv_dynamics_interface_py.h
  )

set(FVDYN_INTERFACE_SRCS
  ${CMAKE_CURRENT_SOURCE_DIR}/fv_dynamics_interface.py
  )

add_custom_command(
  OUTPUT ${FVDYN_INTERFACE_LIBRARY}
  COMMAND ${MPIEXEC_EXECUTABLE} ${MPIEXEC_NUMPROC_FLAG} 1 ${Python_EXECUTABLE} ${FVDYN_INTERFACE_SRCS}
  BYPRODUCTS ${FVDYN_INTERFACE_HEADER_FILE}
  WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
  DEPENDS ${FVDYN_INTERFACE_SRCS}
  COMMENT "Building library with Python"
  VERBATIM
  )

add_custom_target(fv_dynamics_interface_py_h DEPENDS ${FVDYN_INTERFACE_LIBRARY})
add_custom_command(TARGET fv_dynamics_interface_py_h
  POST_BUILD
  COMMAND ${CMAKE_COMMAND} -E copy "${FVDYN_INTERFACE_LIBRARY}" ${CMAKE_INSTALL_PREFIX}/lib)

```

I’m pasting this because maybe an expert will see something bad. For the `INTERFACE` bits:

```cmake
add_library(fv_dynamics_interface_py INTERFACE)
target_include_directories(fv_dynamics_interface_py INTERFACE
  $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
  $<BUILD_INTERFACE:${CMAKE_CURRENT_BINARY_DIR}> # stubs
  # modules and copied *.h, *.inc
  $<BUILD_INTERFACE:${esma_include}/${this}>
  $<INSTALL_INTERFACE:include/${this}>
  )
target_link_libraries(fv_dynamics_interface_py INTERFACE ${FVDYN_INTERFACE_LIBRARY})
add_dependencies(fv_dynamics_interface_py fv_dynamics_interface_py_h)

install(TARGETS fv_dynamics_interface_py
  EXPORT ${PROJECT_NAME}-targets
  LIBRARY DESTINATION ${CMAKE_INSTALL_PREFIX}/lib
  )

```

This is sort of a “borrow from a lot of bits of our code base” hodge-podge of CMake. I’m currently at “Get working” not “make correct” 🙂