better installer behaviour

[Gankra]

Various improvements:

• cargo home
  • respect CARGO_HOME
  • Update ~/.cargo/.crates.toml to register the binary?
• other install dirs
  • install-path customization #235
  • Look for other big likely-on-PATH dirs?
• better installs
  • allow user to customize the installer terminal output #314
  • Run an "install" subcommand of the binary itself?

[Gankra]

Discussing this with Ashley we concluded that it might be for the best if we do like apollo's rover does and try to build out machinery for CLIs to have an install subcommand (which would also theoretically support self-update and uninstalling and whatnot).

That way we can have more hairy logic in rust and not bash scripts (yuck!). This may involve us building out a reusable library for everyone to use for this. It's possible this library already exists in some of the binstalk crates in cargo-binstall, or is aching to be extracted from them.

[Gankra]

Relevant issues rustup had a while back

• Implement shell PATH fixes via UnixShell trait rust-lang/rustup#2387
• Installer modifies ~/.profile which is ignored for graphical terminals rust-lang/rustup#2106

[jyn514]

It would be nice if this also respected CARGO_HOME instead of hardcoding ~/.cargo.

[spitfire05]

Shouldn't the installers install to some more generic destinations than CARGO_HOME, though? After all, if someone's using the installer, there's a high chance he does not have cargo in the first place.

On Unix this can be something like /usr/local/bin/ and some fallback paths. On Windows, it is more tricky, as there are no "default" bin install paths.