Arlo Siemsen

@ehuss could you update the Implementation link to #10907

@Eh2406 I updated this PR to address your comments. I've also cleaned up the test framework to combine the two test http servers (index and api) into one, and added...

### Could we eagerly load config's but lazily load credentials? Cargo currently works by [eagerly loading credentials](https://github.com/rust-lang/cargo/blob/a4c1cd0eb6b18082a7e693f5a665548fe1534be4/src/bin/cargo/commands/publish.rs#L31) for operations that are known to require them (publish, etc). The credentials are...

Thanks for all the feedback. I'll get started on figuring out how to lazy load the `credentials` file. For 1password: * The `op` tool requires a paid 1password account, which...

@ehuss > I think it would be good to keep the CARGO_REGISTRY_NAME_OPT option I've added it back, and used it in the credential process crates > It looks like the...

It's not abandoned. I've been working on writing and implementing [RFC 3289](https://github.com/rust-lang/rfcs/pull/3289) to clean up how source replacement works around authentication. Will be getting back to this soon.

I updated this PR to be based on top of #10907 as two commits.

Using `VSCMD_ARG_VCVARS_SPECTRE` (as this PR does) seems like the safest way to get `rustc` to compile with the spectre mitigated libs. If the user is building within a `spectre` vcvars...

Sure, I can update the PR today. * I'll add the conditional compilation for `not(rust_compiler = "mrustc")` * Looks like the null deref question is resolved now. * `-darwin` is...

I don't have access to the Mac hardware I used when I originally made this change, so I haven't re-validated the full bootstrap of mrustc -> 1.39.0 -> 1.39.0 ->...