cargo package --list requires all dependencies to be published in 1.43.0

[sfackler]

I have some internal distribution packaging logic that uses cargo package --list to find the set of files it should bundle up along with the binary (respecting gitignore, include/exclude, etc).

However, in 1.43 it appears that the command now requires any dependencies of the crate be published for it to work.

Test repository: https://github.com/sfackler/package-regression

On 1.42.0:

$ cargo package --list
warning: manifest has no description, license, license-file, documentation, homepage or repository.
See https://doc.rust-lang.org/cargo/reference/manifest.html#package-metadata for more info.
.cargo_vcs_info.json
.gitignore
Cargo.lock
Cargo.toml
src/main.rs

On 1.43.0:

warning: manifest has no description, license, license-file, documentation, homepage or repository.
See https://doc.rust-lang.org/cargo/reference/manifest.html#package-metadata for more info.
    Updating crates.io index
error: no matching package named `subcrate` found
location searched: registry `https://github.com/rust-lang/crates.io-index`
required by package `package-regression v0.1.0 (/Users/sfackler/package-regression)`

[ehuss]

Yea, the problem is that it now tries to generate Cargo.lock from the package-ready Cargo.toml where the path entries have been removed (here). This was changed in #7905. I'll take a look at deferring that or making it lazy so that it doesn't affect --list.

[sfackler]

If you're poking around, there's also this check that'd be great to remove from --list:

error: all path dependencies must have a version specified when packaging.
dependency `subcrate` does not specify a version.

[alexcrichton]

@sfackler I think that warning/error is addressed by rust-lang/rfcs#2906

[sfackler]

Any chance of folding the fix for this into 1.34.1? cc @Mark-Simulacrum

[Mark-Simulacrum]

I'm happy to do so on release team side, @ehuss or @rust-lang/cargo, can someone sign off on that? I can work on a backport and such tomorrow.

[ehuss]

Seems fine to me. @Mark-Simulacrum or @sfackler are you ok with submitting all the PRs? At a minimum, #8138 will also need to be included to get CI to pass. I think stable will also need #8080, and maybe #8098. There might be others.

[Mark-Simulacrum]

I will try to do the legwork (most likely self approving if no one objects) later today.