Fix rust-src component.

[ehuss]

The rust-src component could not be installed by rustup because it included some symbolic links. #74520 added the backtrace directory which included some symlinks. Since the rust-src component doesn't need most of the files in the backtrace submodule, this changes it to only include the minimum necessary.

Tested with cargo's build-std that it can build from the resulting tarball.

Fixes#74577

[ehuss]

As a followup PR I'd like to look at somehow adding a check so that this particular issue (symlinks) doesn't happen again in the future. I'm not sure where the best place would be to check for that. Maybe in fabricate? Any suggestions?

As for the bigger issue where "rust-src" isn't usable, do you think it is feasible to add a test for that? I'd be reluctant because I can't think of a really stable way to check that. I'm not sure I'd want to gate rust-lang/rust on cargo's build-std feature right now. Maybe there is some other check that would be reasonable? If not, we can probably continue without; it doesn't break too often.

[rbtcollins]

So generally speaking symlinks can't be used on windows without toggling on 'developer mode' - it would be nice to actually remove the symlinks from the tree entirely IMO, though I recognise that that is a larger change.

[Mark-Simulacrum]

@bors r+ p=1

Approving, though I agree this seems like a suboptimal fix in the long run. I think we should consider automatically "delinking" any symlinks we find or something like that.

[bors]

📌 Commit 79673d3 has been approved by Mark-Simulacrum

[bors]

⌛ Testing commit 79673d3 with merge 0b66800a2cfb407c917a83451238298ac22b5245...

[bors]

💔 Test failed - checks-azure

[ehuss]

@bors retry

##[error]The request: 368240 was abandoned due to an infrastructure failure. Notification of assignment to an agent was never received.

macOS x86_64-apple on azure

[bors]

⌛ Testing commit 79673d3 with merge be0c64997c1881943ccfd2220eeef831d12be12e...

[bors]

💥 Test timed out

[ehuss]

@bors retry
Looks like both azure and gha succeeded, but bors didn't get the memo.

[bors]

⌛ Testing commit 79673d3 with merge 4825e12...

[conikeec]

Any update on this PR ? I filed rust-lang/rustup#2434 which in turn is blocked on this. All updates are failing at this point.

[bors]

☀️ Test successful - checks-actions, checks-azure
Approved by: Mark-Simulacrum
Pushing 4825e12 to master...

[RalfJung]

As for the bigger issue where "rust-src" isn't usable, do you think it is feasible to add a test for that? I'd be reluctant because I can't think of a really stable way to check that. I'm not sure I'd want to gate rust-lang/rust on cargo's build-std feature right now. Maybe there is some other check that would be reasonable? If not, we can probably continue without; it doesn't break too often.

x.py test src/tools/miri already uses xargo to build a libstd for Miri. If we could somehow make it use the rust-src component instead of the actual rust source tree, that would provide some pretty good coverage.

[mati865]

x.py test src/tools/miri already uses xargo to build a libstd for Miri. If we could somehow make it use the rust-src component instead of the actual rust source tree, that would provide some pretty good coverage.

If Miri tests don't take too long maybe they could be moved to new builder that starts right when Linux dist finishes. It would install Rust (with proper components) from AWS and then run Miri tests. Don't know if it's possible/feasible though.