Move `download-ci-llvm` out of bootstrap.py
This is ready for review. It has been tested on Windows, Linux, and NixOS.
The second commit ports the changes from https://github.com/rust-lang/rust/pull/95234 to Rust; I can remove it if desired.
Helps with https://github.com/rust-lang/rust/issues/94829.
As a follow-up, this makes it possible to avoid downloading llvm until it's needed for building `rustc_llvm`; it would be nice to do that, but it shouldn't go in the first draft. It might also be possible to avoid requiring python until tests run (currently there's a check in `sanity.rs`), but I haven't looked too much into that.
`@rustbot` label +A-rustbuild
This attempts to keep the logic as close to the original python as possible.
`probably_large` has been removed, since it was always `True`, and UTF-8 paths are no longer supported when patching files for NixOS.
I can readd UTF-8 support if desired.
Note that this required making `llvm_link_shared` computed on-demand,
since we don't know whether it will be static or dynamic until we download LLVM from CI.
Add `x {check,build,doc} {compiler,library}` aliases.
While working on https://github.com/rust-lang/rust/pull/95503, I realized that it will interfere with existing command lines:
Currently people run `x build library/std` expecting it to "add all library crates to the sysroot",
but after that change, it will *only* build `libstd` and its dependencies (and add them to the sysroot), not libtest or libproc_macro.
That will work for local testing in most cases, but could be confusing. Even if not, though, I think `x build library` is more clear about what actually happens than the current `x build library/std`.
The intended end goal is something like:
- For check/build/doc, we have library + compiler aliases, which correspond to basically "most possible" for that piece. This is the intended path of entry (rather than library/test or similar as today) for when you just want the thing to work -- for example, getting a compiler that is "crates.io-compatible" would be roughly `x.py build library`). #95504
- Specific crate invocations build up to that crate, which means that if you don't care about tests you probably want x.py build library/proc_macro or library/std for faster build times. #95503
Note that this is already implemented today for the `doc` command and seems to work pretty well in practice.
I plan to change the dev-guide and various instructions in the README to `build library` once this is merged.
`@rustbot` label +A-rustbuild
These paths (`_cranelift` and `_gcc`) are somewhat misleading, since they
actually tell bootstrap to build *all* codegen backends. But this seems like
a useful improvement in the meantime.
While working on https://github.com/rust-lang/rust/pull/95503,
I realized that this will interfere with existing command lines:
Currently people run `x build library/std` expecting it to be added to the sysroot,
but after that change, it will *only* build `libstd` without making it available
for the toolchain.
It's debatable whether that's a breaking change that will be accepted; if so, this PR is absolutely
necessary to make sure there's a command for "build the standard library and add it to the sysroot".
Even if not, though, I think `x build library` is more clear about what actually happens than the
current `x build library/std`.
For consistency, also add support for `compiler` and all other command variants. Note that `doc
compiler` was already supported, so in a sense this is just fixing an existing inconsistency.
I plan to change the dev-guide and various instructions in the README to `build library` once this is merged.
The majority of the code is only used by either rustbuild or
rustc_llvm's build script. Rust_build is compiled once for rustbuild and
once for every stage. This means that the majority of the code in this
crate is needlessly compiled multiple times. By moving only the code
actually used by the respective crates to rustbuild and rustc_llvm's
build script, this needless duplicate compilation is avoided.
First, this reverts the `CFLAGS`/`CXXFLAGS` of #93918. Those flags are
already read by `cc` and populated into `Build` earlier on in the
process. We shouldn't be overriding that based on `CFLAGS`, since `cc`
also respects overrides like `CFLAGS_{TARGET}` and `HOST_CFLAGS`, which
we want to take into account.
Second, this adds the same capability to specify target-specific
versions of `LDFLAGS` as we have through `cc` for the `C*` flags:
https://github.com/alexcrichton/cc-rs#external-configuration-via-environment-variables
Note that this also necessitated an update to compiletest to treat
CXXFLAGS separately from CFLAGS.
Use the first codegen backend in the config.toml as default
It is currently hard coded to llvm if enabled and cranelift otherwise.
This made some sense when cranelift was the only alternative codegen
backend. Since the introduction of the gcc backend this doesn't make
much sense anymore. Before this PR bootstrapping rustc using a backend
other than llvm or cranelift required changing the source of
rustc_interface. With this PR it becomes a matter of putting the right
backend as first enabled backend in config.toml.
cc ```@antoyo```
It is currently hard coded to llvm if enabled and cranelift otherwise.
This made some sense when cranelift was the only alternative codegen
backend. Since the introduction of the gcc backend this doesn't make
much sense anymore. Before this PR bootstrapping rustc using a backend
other than llvm or cranelift required changing the source of
rustc_interface. With this PR it becomes a matter of putting the right
backend as first enabled backend in config.toml.
Metadata::modified works in all platforms supported by the filetime
crate. This changes brings rustbuild a tiny bit closer towards dropping
the filetime dependency.
Some projects (e.g. the `bootimage` crate) may require the
user to install the `llvm-tools-preview` rustup component.
However, this cannot be easily done with a locally built toolchain.
To allow a local toolchain to be used a drop-in replacement for
a normal rustup toolchain in more cases, this PR copies the built
LLVM tools to the sysoot. From the perspective a tool looking
at the sysroot, this is equivalent to installing `llvm-tools-preview`.
Assemble the compiler when running `x.py build`
Previously, there was no way to actually get binaries in
`build/$TARGET/stage1/bin` without building the standard library. This
makes it possible to build just the compiler. This can be useful when
the standard library isn't actually necessary for trying out your tests
(e.g. a bug that can be reproduced with only a `no_core` crate).
Closes https://github.com/rust-lang/rust/issues/73519.
Previously, there was no way to actually get binaries in
`build/$TARGET/stage1/bin` without building the standard library. This
makes it possible to build just the compiler. This can be useful when
the standard library isn't actually necessary for trying out your tests
(e.g. a bug that can be reproduced with only a `no_core` crate).
The wrapper is installed as `ld` and `ld64` in the `lib\rustlib\<host_target>\bin\gcc-ld`
directory and its sole purpose is to invoke `rust-lld` in the parent directory with
the correct flavor.
Make `-Z gcc-ld=lld` work for Apple targets
`-Z gcc-ld=lld` was introduced in #85961. It does not work on Macos because lld needs be either named `ld64` or passed `-flavor darwin` as the first two arguments in order to select the Mach-O flavor. Rust invokes cc (=clang) on Macos for linking which calls `ld` as linker binary and not `ld64`, so just creating an `ld64` binary and modifying the search path with `-B` does not work.
In order to solve this patch does:
* Set the `lld_flavor` for all Apple-derived targets to `LldFlavor::Ld64`. As far as I can see this actually works towards fixing `-Xlinker=rust-lld` as all those targets use the Mach-O object format.
* Copy/hardlink rust-lld to the gcc-ld subdirectory as ld64 next to ld.
* If `-Z gcc-ld=lld` is used and the target lld flavor is Ld64 add `-fuse-ld=/path/to/ld64` to the linker invocation.
Fixes#86945.
This only updates the submodules the first time they're needed, instead
of unconditionally the first time you run x.py.
Ideally, this would move *all* submodules and not exclude some tools and
backtrace. Unfortunately, cargo requires all `Cargo.toml` files in the
whole workspace to be present to build any crate.
On my machine, this takes the time for an initial submodule clone (for
`x.py --help`) from 55.70 to 15.87 seconds.
This uses exactly the same logic as the LLVM update used, modulo some
minor cleanups:
- Use a local variable for `src.join(relative_path)`
- Remove unnecessary arrays for `book!` macro and make the macro simpler to use
- Add more comments
- Don't print the exact command run by rustbuild unless `--verbose` is set.
This is almost always unhelpful, since it's just cargo with a lot of
arguments.
- Don't print "Build completed unsuccessfully" unless --verbose is set.
You can already tell the build failed by the errors above, and the
time isn't particularly helpful.
- Don't print the full path to bootstrap. This is useless to everyone,
even including when working on x.py itself. You can still opt-in to
this being shown with `--verbose`, since it will throw an exception.
Before:
```
error[E0432]: unresolved import `x`
--> library/std/src/lib.rs:343:5
|
343 | use x;
| ^ no external crate `x`
error: aborting due to previous error
For more information about this error, try `rustc --explain E0432`.
error: could not compile `std`
To learn more, run the command again with --verbose.
command did not execute successfully: "/home/joshua/rustc4/build/x86_64-unknown-linux-gnu/stage0/bin/cargo" "check" "--target" "x86_64-unknown-linux-gnu" "-Zbinary-dep-depinfo" "-j" "8" "--release" "--features" "panic-unwind backtrace" "--manifest-path" "/home/joshua/rustc4/library/test/Cargo.toml" "--message-format" "json-render-diagnostics"
expected success, got: exit status: 101
failed to run: /home/joshua/rustc4/build/bootstrap/debug/bootstrap check
Build completed unsuccessfully in 0:00:13
```
After:
```
error[E0432]: unresolved import `x`
--> library/std/src/lib.rs:343:5
|
343 | use x;
| ^ no external crate `x`
error: aborting due to previous error
For more information about this error, try `rustc --explain E0432`.
error: could not compile `std`
To learn more, run the command again with --verbose.
```
ignore test if rust-lld not found
create ld -> rust-lld symlink at build time instead of run time
for testing in ci
copy instead of symlinking
remove linux check
test for linker, suggestions from bjorn3
fix overly restrictive lld matcher
use -Zgcc-ld flag instead of -Clinker-flavor
refactor code adding lld to gcc path
revert ci changes
suggestions from petrochenkov
rename gcc_ld to gcc-ld in dirs
Remove the install prefix from the rpath set when using -Crpath
It was broken anyway for rustup installs and nobody seems to have noticed.
Fixes https://github.com/rust-lang/rust/issues/82392
BPF target support
This adds `bpfel-unknown-none` and `bpfeb-unknown-none`, two new no_std targets that generate little and big endian BPF. The approach taken is very similar to the cuda target, where `TargetOptions::obj_is_bitcode` is enabled and code generation is done by the linker.
I added the targets to `dist-various-2`. There are [some tests](https://github.com/alessandrod/bpf-linker/tree/main/tests/assembly) in bpf-linker and I'm planning to add more. Those are currently not ran as part of rust CI.
## User-facing changes
- Intra-doc links to primitives that currently go to rust-lang.org/nightly/std/primitive.x.html will start going to channel that rustdoc was built with. Nightly will continue going to /nightly; Beta will link to /beta; stable compilers will link to /1.52.1 (or whatever version they were built as).
- Cross-crate links from std to core currently go to /nightly unconditionally. They will start going to /1.52.0 on stable channels (but remain the same on nightly channels).
- Intra-crate links from std to std (or core to core) currently go to the same URL they are hosted at; they will continue to do so. Notably, this is different from everything else because it can preserve the distinction between /stable and /1.52.0 by using relative links.
Note that "links" includes both intra-doc links and rustdoc's own
automatically generated hyperlinks.
## Implementation changes
- Update the testsuite to allow linking to /beta and /1.52.1 in docs
- Use an html_root_url for the standard library that's dependent on the channel
This avoids linking to nightly docs on stable.
- Update rustdoc to use channel-dependent links for primitives from an
unknown crate
- Set DOC_RUST_LANG_ORG_CHANNEL from bootstrap to ensure it's in sync
- Include doc.rust-lang.org in the channel
## Motivation
This avoids having to rebuild bootstrap and tidy each time you rebase
over master. In particular, it makes rebasing and running `x.py fmt` on
each commit in a branch significantly faster. It also avoids having to
rebuild bootstrap after setting `download-rustc = true`.
## Implementation
Instead of extracting the CI artifacts directly to `stage0/`, extract
them to `ci-rustc/` instead. Continue to copy them to the proper
sysroots as necessary for all stages except stage 0.
This also requires `bootstrap.py` to download both stage0 and CI
artifacts and distinguish between the two when checking stamp files.
Note that since tools have to be built by the same compiler that built
`rustc-dev` and the standard library, the downloaded artifacts can't be
reused when building with the beta compiler. To make sure this is still
a good user experience, warn when building with the beta compiler, and
default to building with stage 2.
This switches Rust's WASI target to use crt1-command.o instead of
crt1.o, which enables support for new-style commands. By default,
new-style commands work the same way as old-style commands, so nothing
immediately changes here, but this will be needed by later changes to
enable support for typed arguments.
See here for more information on new-style commands:
- https://github.com/WebAssembly/wasi-libc/pull/203
- https://reviews.llvm.org/D81689
Remove `ENABLE_DOWNLOAD_RUSTC` constant
`ENABLE_DOWNLOAD_RUSTC` was introduced as part of the MVP for `download-rustc` as a way not to rebuild artifacts that have already been downloaded. Unfortunately, it doesn't work very well:
- Steps are ignored by default, which makes it easy to leave out a step
that should be built. For example, the MVP forgot to enable any tests,
so it was only possible to *build* locally.
- It didn't work correctly even when it was enabled: calling
`builder.ensure()` would completely ignore the constant and rebuild the
step anyway. This has no obvious fix since `ensure()` has to return a
`Step::Output`.
Instead, this handles `download-rustc` in `impl Step for Rustc` and
`impl Step for Std`, which to my knowledge are the only build steps that
don't first go through `impl Step for Sysroot` (`Rustc` is used for
the `rustc-dev` component).
See https://github.com/rust-lang/rust/pull/79540#discussion_r563350075 and https://github.com/rust-lang/rust/issues/81930 for further context.
Here are some example runs with these changes and `download-rustc`
enabled:
```
$ x.py build src/tools/clippy
Building stage1 tool clippy-driver (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 1m 09s
Building stage1 tool cargo-clippy (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 0.11s
$ x.py test src/tools/clippy
Finished dev [unoptimized + debuginfo] target(s) in 0.09s
Building stage1 tool clippy-driver (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 0.09s
Building rustdoc for stage1 (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 0.28s
Finished release [optimized] target(s) in 15.26s
Running build/x86_64-unknown-linux-gnu/stage1-tools/x86_64-unknown-linux-gnu/release/deps/clippy_driver-8b407b140e0aa91c
test result: ok. 592 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out
$ x.py build src/tools/rustdoc
Building rustdoc for stage1 (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 41.28s
Build completed successfully in 0:00:41
$ x.py test src/test/rustdoc-ui
Building stage0 tool compiletest (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 0.12s
Building rustdoc for stage1 (x86_64-unknown-linux-gnu)
Finished release [optimized] target(s) in 0.10s
test result: ok. 105 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 8.15s
$ x.py build compiler/rustc
Finished dev [unoptimized + debuginfo] target(s) in 0.09s
Build completed successfully in 0:00:00
```
Note a few things:
- Clippy depends on stage1 rustc-dev artifacts, but rustc didn't have to
be recompiled. Instead, the artifacts were copied automatically.
- All steps are always enabled. There is no danger of forgetting a step,
since only the entrypoints have to handle `download-rustc`.
- Building the compiler (`compiler/rustc`) automatically does no work.
Helps with https://github.com/rust-lang/rust/issues/81930.
r? `@Mark-Simulacrum`
Improve error messages for musl-libdir and wasi-root keys. Previously
the parser would panic with `unwrap()`. Now it prints
Target "wasm32-wasi" does not have a "wasi-root" key
(and similar for the `musl-libdir` field, which is used in target that
use musl)
Also update comments around wasi-root field to make it clear that the
field is only valid in wasm32-wasi target and needs to be moved to a
`[target.wasm32-wasi]` section to be valid.
Fixes#82317
For some targets, rustc uses a "CRT fallback", where it links CRT
object files it ships instead of letting the host compiler link
them.
On musl, rustc currently links crt1, crti and crtn (provided by
libc), but does not link crtbegin and crtend (provided by libgcc).
In particular, crtend is responsible for terminating the .eh_frame
section. Lack of terminator may result in segfaults during
unwinding, as reported in #47551 and encountered by the LLVM 12
update in #81451.
This patch links crtbegin and crtend for musl as well, following
the table at the top of crt_objects.rs.
