make: build virtme-ng-init by default

[matttbe]

As mentioned in the README file, the make command should only build virtme-ng-init. Building is what make usually does, and not installing stuff.

Since commit 85a9b42 ("Make it build on macOS"), make was building virtme-ng-init, but it was also trying to install it with other files elsewhere on the system, what make install does. That's because make was using the first target as the default one, so the install one.

Instead of directly calling cargo like it was the case before, the build part of setup.py is called. This will generate files in the build dir that are probably not needed, but it is quick, and it will also build virtme-ng-init properly, what we want. Cc: @tamird.

While at it:

• Mention all targets that are not linked to files in .PHONY.

• Only generate the init file when needed: when the .rs files have been modified.

• Initialise the submodule if the .rs files are not present.

• A new clean target can help to force rebuilding (or just to clean, even if it looks like it missed many generated files...).

[tamird]

Since commit 85a9b42 ("Make it build on macOS"), make was building virtme-ng-init, but it was also trying to install it, what make install does.

Are you referring to the use of cargo install? Because that isn't "installing" in the way you're implying because of the --root argument which causes cargo to "install" into the package, not system-wide destination.

I'm not really sure what problem this PR is solving but this description is misleading. Please consider describing the true motivation.

[matttbe]

Since commit 85a9b42 ("Make it build on macOS"), make was building virtme-ng-init, but it was also trying to install it, what make install does.

Are you referring to the use of cargo install? Because that isn't "installing" in the way you're implying because of the --root argument which causes cargo to "install" into the package, not system-wide destination.

I'm not really sure what problem this PR is solving but this description is misleading. Please consider describing the true motivation.

Thanks, I just updated the description: we don't want make to run make install by default, just the build for a local use:

Since commit 85a9b42 ("Make it build on macOS"), make was building virtme-ng-init, but it was also trying to install it with other files elsewhere on the system, what make install does. That's because make was using the first target as the default one, so the install one.

Is it clearer?

[matttbe]

Note: I used python3 setup.py build, but maybe there is a better command to use?

[arighi]

Note: I used python3 setup.py build, but maybe there is a better command to use?

We could detect python3 || python, but I wouldn't over-complicate things here, running python3 setup.py build seems good enough.

[tamird]

Thanks for clarifying.

[matttbe]

Note: I used python3 setup.py build, but maybe there is a better command to use?

We could detect python3 || python, but I wouldn't over-complicate things here, running python3 setup.py build seems good enough.

Good point. We can indeed change that later. Are there still some distributions not having a python3 binary, only python? I remembered issues a few years ago when using python, but not sure about python3.

[arighi]

Good point. We can indeed change that later. Are there still some distributions not having a python3 binary, only python? I remembered issues a few years ago when using python, but not sure about python3.

Not that I'm aware of, I think all the recent distro have python3.