r? @alexcrichton

(rust_highfive has picked a reviewer for you, use r? to override)

Oh, Tests failed.
I'll look into that.

Thanks! Could you be sure to add some tests for the new functionality as well?

That should have fixed the tests and also simplified the changes.
I could not check the cross-compiling-tests though.

Ok, got a chance to look at this. Cleared up on IRC that new new functionality was added to the CLI, just the backend.

Unfortunately though I think this suffers the same problem as #2601 where once the support is hooked up it'll print out and update the registry one time per crate being installed. That experience though, I believe, was the primary stalling point for the previous PR.

Maybe it'd be good to go ahead and add support to the CLI all in one go to see how it'd work?

This is indeed a problem.
I looked into the code for a bit.

It seems to me that we want to prevent the Registry::update function from being triggered multiple times.

Also, I'm not yet sure how to parse the arguments.
Docopt is probably not sufficient for what we would want to do. I asked some people and also opened an issue ( docopt/docopt#344 ), but I'm almost certain this cannot be done with it.

I might look into clap or something similar.

Yeah I think we'll basically want to persist the same Registry across multiple installations so it can retain the state of what's updated. I'd also prefer to avoid two command line parsers for now, let's stick to docopt and try to fix bugs and/or work around it.

I agree to having a persistent Registry.
But I think that it is not possible to implement a proper command parsing solution solely with docopt.

☔ The latest upstream changes (presumably #3146) made this pull request unmergeable. Please resolve the merge conflicts.

Closing due to inactivity, but feel free to resubmit with a rebase!