bootstrap: copy self-contained linking components to stage0-sysroot

[lqd]

I hit this issue while trying to bootstrap using a rustc where rust-lld is used by default: this was the cause of the failure to profile rustc-perf's bootstrap benchmark in #113382.

stage0-sysroot currently only has libs and self-contained objects, not the other self-contained linking components yet. Most notably, it does not contain the linker and wrappers that we build, and that rustup distributes.

If you try to bootstrap using the bootstrap compiler's rust-lld, it will fail to link std at stage0 because rust-lld and the gcc-ld wrappers, will not be found in stage0-sysroot/lib/rustlib/x86_64-unknown-linux-gnu/bin.

This PR copies the bin directory next to the lib directory when rust.lld is enabled in the config (though maybe it could be done unconditionally, the fact that we need it to link does not necessarily mean that we'd want to build and provide it at stage1).

cc @Kobzol who also encountered this issue while using lld during bootstrap.

[rustbot]

r? @onur-ozkan

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

[Kobzol]

I think that it should be done unconditionally. We should copy the things required by the snapshot compiler to stage0-sysroot, so that it can Do its thing™. Especially if it will eventually use rust-lld by default, because in that case rust-lld needs to be there, so that rustup link stage0 ... continues to work.

rust.lld = true specifies whether lld should appear in stage1/2/N compiler's sysroot, which is an orthogonal decision, IMO.

[onur-ozkan]

Doing it unconditionally seems better in my opinion too. r=me with doing it unconditionally.

[lqd]

I've removed the rust.lld check, making the copy unconditional.

@bors r=onur-ozkan

[bors]

📌 Commit 5fde111 has been approved by onur-ozkan

It is now in the queue for this repository.

[bors]

⌛ Testing commit 5fde111 with merge 2685d3e...

[bors]

💔 Test failed - checks-actions

[lqd]

Evidently the snapshot sysroot libdir is "bin" on windows and not "lib", pointing to a non-existent path to rust-lld.
I've pushed an update locating the source sysroot bin folder from the root of the snapshot sysroot, and avoid this bin/lib difference.

i.e. not using rustc_snapshot_libdir since that is "bin" on windows, and not the needed rustc_snapshot_sysroot() + "/lib": the path to rust-lld is always lib/rustlib/host/bin.

[onur-ozkan]

@bors r+

[bors]

📌 Commit 82d23a6 has been approved by onur-ozkan

It is now in the queue for this repository.

[bors]

⌛ Testing commit 82d23a6 with merge b0889cb...

[bors]

☀️ Test successful - checks-actions
Approved by: onur-ozkan
Pushing b0889cb to master...

[rust-timer]

Finished benchmarking commit (b0889cb): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-4.5%	[-4.5%, -4.5%]	1
All ❌✅ (primary)	-	-	0

Cycles

This benchmark run did not return any relevant results for this metric.

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 626.682s -> 628.193s (0.24%)
Artifact size: 273.33 MiB -> 273.37 MiB (0.02%)

[tmandry]

Our stage0 compiler doesn't include rust-lld because we don't use it to link. Shouldn't use-lld tell you whether lld is needed or not?

The unconditional copy causes the build to fail. Can this be reverted? EDIT: #116365 should fix this

[lqd]

Shouldn't use-lld tell you whether lld is needed or not?

Not really no, use-lld uses system lld (unless on msvc maybe).

[tmandry]

That's not what I interpret config.example.toml to mean:

# Indicates whether LLD will be used to link Rust crates during bootstrap on
# supported platforms. The LLD from the bootstrap distribution will be used
# and not the LLD compiled during the bootstrap.

"LLD from the bootstrap distribution" sounds pretty clearly like stage0. Is the comment wrong?

[lqd]

Is the comment wrong?

That seems to be the case: the most recent look at lld usage in bootstrap, #116278, describes this as being maybe true only on msvc.

[tmandry]

Gah, that's annoying. Glad we have people cleaning this up now though!

[Kobzol]

I guess that the copy should be unconditional of the contents of config.toml, but conditional of the actual presence of rust-lld in the stage0 compiler.

[lqd]

Agreed. That’s what I suggested in the new PR.