tools no longer find librustc_driver

[matthiaskrgr]

~/.cargo/bin/cargo +master fmt

/home/matthias/.rustup/toolchains/master/bin/rustfmt: error while loading shared libraries: librustc_driver-db25bd44b4d5bdcc.so: cannot open shared object file: No such file or directory

~/.cargo/bin/cargo +master clippy

/home/matthias/.rustup/toolchains/master/bin/cargo-clippy: error while loading shared libraries: librustc_driver-db25bd44b4d5bdcc.so: cannot open shared object file: No such file or directory

rustc 1.87.0-nightly (bb2cc59a2 2025-02-23)
binary: rustc
commit-hash: bb2cc59a2172d6e35c89b409a4e6b5058d9039d7
commit-date: 2025-02-23
host: x86_64-unknown-linux-gnu
release: 1.87.0-nightly
LLVM version: 20.1.0

cc @onur-ozkan #137215

[onur-ozkan]

I will check it today

[onur-ozkan]

I tried this: x build compiler std cargo clippy rustfmt

and they seem to be working:

$  ~/devspace/.other/rustc-builds  $ ls ./build/host/stage1/bin
cargo-clippy  cargo-fmt  clippy-driver  rustc  rustfmt
$  ~/devspace/.other/rustc-builds  $ ./build/host/stage1/bin/cargo-fmt --version
rustfmt 1.8.0-dev (bb2cc59a21 2025-02-23)
$  ~/devspace/.other/rustc-builds  $ ./build/host/stage1/bin/rustfmt --version
rustfmt 1.8.0-dev (bb2cc59a21 2025-02-23)
$  ~/devspace/.other/rustc-builds  $ ./build/host/stage1/bin/cargo-clippy --version
clippy 0.1.86 (bb2cc59a21 2025-02-23)

How did you perform your test?

[matthiaskrgr]

with https://github.com/kennytm/rustup-toolchain-install-master I do a

rustup-toolchain-install-master -f -n master -c rustc-dev llvm-tools rust-src clippy rust-analyzer rustfmt miri rustc-codegen-cranelift rust-docs
(installs master toolchain with a bunch of components)

and then
~/.cargo/bin/cargo +master clippy
to run this master toolchain via cargo

[Noratrieb]

nora@nixos /t/t/meow (main)> readelf -d /home/nora/.rustup/toolchains/master/bin/rustfmt

Dynamic section at offset 0x50b6c8 contains 32 entries:
  Tag        Type                         Name/Value
 0x000000000000001d (RUNPATH)            Library runpath: [$ORIGIN/../lib]
 0x0000000000000001 (NEEDED)             Shared library: [librustc_driver-db25bd44b4d5bdcc.so]

It has the rpath and NEEDED, so the tool binary itself looks fine.

nora@nixos /t/t/meow (main)> ll /home/nora/.rustup/toolchains/master/lib/librustc*
-rw-r--r-- 1 nora users 141684152 Feb 23 11:06 /home/nora/.rustup/toolchains/master/lib/librustc_driver-19054624313a00ee.so

But the librustc_driver has a different hash..... somehow the tools are built against the wrong rustc.

[Noratrieb]

This feels related to #123384, maybe that tool staging bug messed up your new build logic?

[onur-ozkan]

$  ~/devspace/.other/rustc-builds  $ readelf -d build/host/stage2/bin/rustdoc

Dynamic section at offset 0xc051b0 contains 32 entries:
  Tag        Type                         Name/Value
 0x0000000000000001 (NEEDED)             Shared library: [librustc_driver-e09e126965159615.so]
 0x0000000000000001 (NEEDED)             Shared library: [libm.so.6]
 0x0000000000000001 (NEEDED)             Shared library: [libgcc_s.so.1]
 0x0000000000000001 (NEEDED)             Shared library: [libc.so.6]
 0x0000000000000001 (NEEDED)             Shared library: [ld-linux-x86-64.so.2]
 0x000000000000001d (RUNPATH)            Library runpath: [$ORIGIN/../lib]
$  ~/devspace/.other/rustc-builds  $ ls build/host/stage2/lib
librustc_driver-e09e126965159615.so  rustlib

Regular builds seems fine, will look into x dist results.

[onur-ozkan]

Yeah, dist logic doesn't seem to be right... Working on the fix now.

[matthiaskrgr]

I wonder if we can have a little smoke test on CI

1. create dist artifacts
2. install them via rustup
3. cargo new testcrate
4. cd testcrate
5. cargo +dist build
6. cargo +dist fmt
7. cargo +dist clippy
8. cargo +dist doc

or something like that

[Noratrieb]

ohhhhh, your #137215 did #123384, that's very cool! nice.

@matthiaskrgr yeah, having dist-smoke-tests would be useful. see #136822 when recently i686-pc-windows-gnu suddenly had x86-64 libraries :D

[Noratrieb]

I've prepared a revert in #137475 in case we don't find a fix for the bug in time, so there's no rush :).

[onur-ozkan]

I found the bug and did the fix, just preparing the PR now :)