the fix does not make sense to me. install_in_hostpython = True is a directive to build the recipe twice, once for target arch, and once for host arch. You would only need to build for hostpython if some other recipe has a build dependency on tflite-runtime, which to my knowledge, is not the cause of issue #2630.

Many recipes dont have this 'install_in_hostpython' directive, and do not exhibit issue #2630 .

Am I missing something?

the fix does not make sense to me. install_in_hostpython = True is a directive to build the recipe twice, once for target arch, and once for host arch. You would only need to build for hostpython if some other recipe has a build dependency on tflite-runtime, which to my knowledge, is not the cause of issue #2630.

Many recipes dont have this 'install_in_hostpython' directive, and do not exhibit issue #2630 .

Am I missing something?

It's really strange, but you can check if for some reason the package stops being reassembled

Agree, what should be looked into instead is probably the PythonRecipe.should_build().
You can override this function in your recipe, put a breakpoint and step into super().should_build(arch).
My guess would be there's some kind of mismatch between the recipe name and what's to be expected in that folder, well something like that.
But that requires a bit of digging

Agree, what should be looked into instead is probably the PythonRecipe.should_build(). You can override this function in your recipe, put a breakpoint and step into super().should_build(arch). My guess would be there's some kind of mismatch between the recipe name and what's to be expected in that folder, well something like that. But that requires a bit of digging

Done

Hi @Neizvestnyj

Seems that the PEP8 tests are failing (and so also the other tests), can you take care of it?

Done