[beta] rustbuild: set __CARGO_DEFAULT_LIB_METADATA to channel

[Keruspe]

This is a backport of #42977 for beta

[rust-highfive]

r? @nikomatsakis

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

[Keruspe]

r? @alexcrichton

[Mark-Simulacrum]

I think this needs to update Cargo.lock as well? Something like cargo update -p https://github.com/rust-lang/cargo#0.21.0 --manifest-path src/Cargo.toml should work, with a different version number perhaps.

[Keruspe]

@Mark-Simulacrum Out of curiosity, what would require such a change? There's no Cargo.toml change in this pull request

[Mark-Simulacrum]

Right, but if you look at the Cargo.lock, https://github.com/rust-lang/rust/blob/beta/src/Cargo.lock#L149, the git commit hash is included. Cargo.toml has https://github.com/rust-lang/rust/blob/beta/src/Cargo.toml#L39, which is version-based, which is why it doesn't need an update.

[Mark-Simulacrum]

It's possible that in practice, this doesn't matter, since we use the submodule for Cargo and (perhaps?) Cargo ignores the commit hash in the submodule when utilizing the Cargo from there. The PR builder doesn't actually build the "extended" release so it's hard to check what happens based on that...

[Keruspe]

Interresting.
Fwiw, i run the extended build locally and it went just fine (and doesn't seem to have updated Cargo.lock in the process)

[Mark-Simulacrum]

@alexcrichton Do you know if Cargo currently checks the git commit hash from Cargo.lock against submodules? It feels based on the above that it doesn't, which seems somewhat unfortunate...

[alexcrichton]

@bors: r+

@Mark-Simulacrum yeah I think it's definitely clear there's a bug/downside or two to how we redirect the rls to depend on the Cargo in-tree ...

[bors]

📌 Commit 427cbbc has been approved by alexcrichton

[bors]

⌛ Testing commit 427cbbc with merge c064c87...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: alexcrichton
Pushing c064c87 to beta...