Failure when ./install_requirements.sh: cannot find -lextension_tensor

[Jack-Khuu]

🐛 Describe the bug

After a pinbump in torchchat of ET (2025-01-16) pytorch/torchchat#1459, one of our CI tests started to fail when installing executorch

Example: https://github.com/pytorch/torchchat/actions/runs/12916656365/job/36021302921

Specifically the error encountered is: /opt/rh/gcc-toolset-11/root/usr/bin/ld: cannot find -lextension_tensor

 [100%] Linking CXX shared library _portable_lib.cpython-311-x86_64-linux-gnu.so
  /opt/rh/gcc-toolset-11/root/usr/bin/ld: cannot find -lextension_tensor
  collect2: error: ld returned 1 exit status
  gmake[3]: *** [CMakeFiles/portable_lib.dir/build.make:119: _portable_lib.cpython-311-x86_64-linux-gnu.so] Error 1
  gmake[2]: *** [CMakeFiles/Makefile2:752: CMakeFiles/portable_lib.dir/all] Error 2
  gmake[1]: *** [CMakeFiles/Makefile2:759: CMakeFiles/portable_lib.dir/rule] Error 2
  gmake: *** [Makefile:208: portable_lib] Error 2
  error: command '/opt/conda/bin/cmake' failed with exit code 2
  error: subprocess-exited-with-error
  
  × Building wheel for executorch (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> See above for output.

Is there a specific NDK or CMAKE arg required to pick the extension_tensors up?

Versions

9c04329

[Jack-Khuu]

Bumping to NDK to android-ndk-r27c-linux.zip doesn't fix, but probably isn't wrong either cc: @kirklandsign

PR: pytorch/torchchat#1473

[Jack-Khuu]

Resolved in pytorch/torchchat#1475

Had to update ET install instructions. #7473 enabled XNN pybindings on by default