These are quite long, usually, and in most cases not interesting. On smaller
terminals they can take up more than a full page of output, hiding the error
diagnostics emitted.
The directories for core, alloc, std, proc_macro, and test crates now
correspond directly to the crate name and stripping the "lib" prefix is
no longer necessary.
This commit is a proof-of-concept for switching the standard library's
backtrace symbolication mechanism on most platforms from libbacktrace to
gimli. The standard library's support for `RUST_BACKTRACE=1` requires
in-process parsing of object files and DWARF debug information to
interpret it and print the filename/line number of stack frames as part
of a backtrace.
Historically this support in the standard library has come from a
library called "libbacktrace". The libbacktrace library seems to have
been extracted from gcc at some point and is written in C. We've had a
lot of issues with libbacktrace over time, unfortunately, though. The
library does not appear to be actively maintained since we've had
patches sit for months-to-years without comments. We have discovered a
good number of soundness issues with the library itself, both when
parsing valid DWARF as well as invalid DWARF. This is enough of an issue
that the libs team has previously decided that we cannot feed untrusted
inputs to libbacktrace. This also doesn't take into account the
portability of libbacktrace which has been difficult to manage and
maintain over time. While possible there are lots of exceptions and it's
the main C dependency of the standard library right now.
For years it's been the desire to switch over to a Rust-based solution
for symbolicating backtraces. It's been assumed that we'll be using the
Gimli family of crates for this purpose, which are targeted at safely
and efficiently parsing DWARF debug information. I've been working
recently to shore up the Gimli support in the `backtrace` crate. As of a
few weeks ago the `backtrace` crate, by default, uses Gimli when loaded
from crates.io. This transition has gone well enough that I figured it
was time to start talking seriously about this change to the standard
library.
This commit is a preview of what's probably the best way to integrate
the `backtrace` crate into the standard library with the Gimli feature
turned on. While today it's used as a crates.io dependency, this commit
switches the `backtrace` crate to a submodule of this repository which
will need to be updated manually. This is not done lightly, but is
thought to be the best solution. The primary reason for this is that the
`backtrace` crate needs to do some pretty nontrivial filesystem
interactions to locate debug information. Working without `std::fs` is
not an option, and while it might be possible to do some sort of
trait-based solution when prototyped it was found to be too unergonomic.
Using a submodule allows the `backtrace` crate to build as a submodule
of the `std` crate itself, enabling it to use `std::fs` and such.
Otherwise this adds new dependencies to the standard library. This step
requires extra attention because this means that these crates are now
going to be included with all Rust programs by default. It's important
to note, however, that we're already shipping libbacktrace with all Rust
programs by default and it has a bunch of C code implementing all of
this internally anyway, so we're basically already switching
already-shipping functionality to Rust from C.
* `object` - this crate is used to parse object file headers and
contents. Very low-level support is used from this crate and almost
all of it is disabled. Largely we're just using struct definitions as
well as convenience methods internally to read bytes and such.
* `addr2line` - this is the main meat of the implementation for
symbolication. This crate depends on `gimli` for DWARF parsing and
then provides interfaces needed by the `backtrace` crate to turn an
address into a filename / line number. This crate is actually pretty
small (fits in a single file almost!) and mirrors most of what
`dwarf.c` does for libbacktrace.
* `miniz_oxide` - the libbacktrace crate transparently handles
compressed debug information which is compressed with zlib. This crate
is used to decompress compressed debug sections.
* `gimli` - not actually used directly, but a dependency of `addr2line`.
* `adler32`- not used directly either, but a dependency of
`miniz_oxide`.
The goal of this change is to improve the safety of backtrace
symbolication in the standard library, especially in the face of
possibly malformed DWARF debug information. Even to this day we're still
seeing segfaults in libbacktrace which could possibly become security
vulnerabilities. This change should almost entirely eliminate this
possibility whilc also paving the way forward to adding more features
like split debug information.
Some references for those interested are:
* Original addition of libbacktrace - #12602
* OOM with libbacktrace - #24231
* Backtrace failure due to use of uninitialized value - #28447
* Possibility to feed untrusted data to libbacktrace - #21889
* Soundness fix for libbacktrace - #33729
* Crash in libbacktrace - #39468
* Support for macOS, never merged - ianlancetaylor/libbacktrace#2
* Performance issues with libbacktrace - #29293, #37477
* Update procedure is quite complicated due to how many patches we
need to carry - #50955
* Libbacktrace doesn't work on MinGW with dynamic libs - #71060
* Segfault in libbacktrace on macOS - #71397
Switching to Rust will not make us immune to all of these issues. The
crashes are expected to go away, but correctness and performance may
still have bugs arise. The gimli and `backtrace` crates, however, are
actively maintained unlike libbacktrace, so this should enable us to at
least efficiently apply fixes as situations come up.
Improve defaults in x.py
- Make the default stage dependent on the subcommand
- Don't build stage1 rustc artifacts with x.py build --stage 1. If this is what you want, use x.py build --stage 2 instead, which gives you a working libstd.
- Change default debuginfo when debug = true from 2 to 1
I tried to fix CI to use `--stage 2` everywhere it currently has no stage, but I might have missed a spot.
This does not update much of the documentation - most of it is in https://github.com/rust-lang/rustc-dev-guide/ or https://github.com/rust-lang/rust-forge and will need a separate PR.
See individual commits for a detailed rationale of each change.
See also the MCP: https://github.com/rust-lang/compiler-team/issues/326
r? @Mark-Simulacrum , but anyone is free to give an opinion.
- Use stage 2 for makefile
- Move assert to builder
- Don't add an assert for --help
- Allow --stage 0 if passed explicitly
- Don't assert defaults during tests
Otherwise it's impossible to test the defaults!
From [a conversation in discord](https://discordapp.com/channels/442252698964721669/443151243398086667/719200989269327882):
> Linking seems to consume all available RAM, leading to the OS to swap memory to disk and slowing down everything in the process
Compiling itself doesn't seem to take up as much RAM, and I'm only looking to check whether a minimal testcase can be compiled by rustc, where the runtime performance isn't much of an issue
> do you have debug = true or debuginfo-level = 2 in config.toml?
> if so I think that results in over 2GB of debuginfo nowadays and is likely the culprit
> which might mean we're giving out bad advice :(
Anecdotally, this sped up my stage 1 build from 15 to 10 minutes.
This still adds line numbers, it only removes variable and type information.
- Improve wording for debuginfo description
Co-authored-by: Teymour Aldridge <42674621+teymour-aldridge@users.noreply.github.com>
- Set rustc to build only when explicitly asked for
This allows building the stage2 rustc artifacts, which nothing depends
on.
Previously the behavior was as follows (where stageN <-> stage(N-1) artifacts, except for stage0 libstd):
- `x.py build --stage 0`:
- stage0 libstd
- stage1 rustc (but without putting rustc in stage0/)
This leaves you without any rustc at all except for the beta compiler
(https://github.com/rust-lang/rust/issues/73519). This is never what you want.
- `x.py build --stage 1`:
- stage0 libstd
- stage1 rustc
- stage1 libstd
- stage1 rustdoc
- stage2 rustc
This leaves you with a broken stage2 rustc which doesn't even have
libcore and is effectively useless. Additionally, it compiles rustc
twice, which is not normally what you want.
- `x.py build --stage 2`:
- stage0 libstd
- stage1 rustc
- stage1 libstd
- stage2 rustc
- stage2 rustdoc and tools
This builds all tools in release mode. This is the correct usage for CI,
but takes far to long for development.
Now the behavior is as follows:
- `x.py build --stage 0`:
- stage0 libstd
This is suitable for contributors only working on the standard library,
as it means rustc never has to be compiled.
- `x.py build --stage 1`:
- stage0 libstd
- stage1 rustc
- stage1 libstd
- stage1 rustdoc
This is suitable for contributors working on the compiler. It ensures
that you have a working rustc and libstd without having to pass
`src/libstd` in addition.
- `x.py build --stage 2`:
- stage0 libstd
- stage1 rustc
- stage1 libstd
- stage2 rustc
- stage2 libstd
- stage2 rustdoc
This is suitable for debugging errors which only appear with the stage2
compiler.
- `x.py build --stage 2 src/libstd src/rustc`
- stage0 libstd
- stage1 rustc
- stage1 libstd
- stage2 rustc
- stage2 libstd
- stage2 rustdoc, tools, etc.
- stage2 rustc artifacts ('stage3')
This is suitable for CI, which wants all tools in release mode.
However, most of the use cases for this should use `x.py dist` instead,
which builds all the tools without each having to be named individually.
### x.py build/test: stage 1
I've seen very few people who actually use full stage 2 builds on purpose. These compile rustc and libstd twice and don't give you much more information than a stage 1 build (except in rare cases like https://github.com/rust-lang/rust/pull/68692#discussion_r376392145). For new contributors, this makes the build process even more daunting than it already is. As long as CI is changed to use `--stage 2` I see no downside here.
### x.py bench/dist/install: stage 2
These commands have to do with a finished, optimized version of rustc. It seems very rare to want to use these with a stage 1 build.
### x.py doc: stage 0
Normally when you document things you're just fixing a typo. In this case there is no need to build the whole rust compiler, since the documentation will usually be the same when generated with the beta compiler or with stage 1.
Note that for this release cycle only there will be a significant different between stage0 and stage1 docs: https://github.com/rust-lang/rust/pull/73101. However most of the time this will not be the case.
Revert libbacktrace -> gimli
This reverts 4cbd265c11028f8d7b8513db3cc1e8d7a36d8964 (and technically 79673d3009 but it's made empty by previous reverts).
The current plan is to land this PR as a temporary change, so that we can get a better handle on the regressions introduced by it. Trying to fix/examine them in master is difficult, and we want to be better able to evaluate them without impact to other PRs being landed in the mean time.
That said, it is currently *my* belief that gimli, in one form or another, will need to land sometime soon. I think it's quite likely that it may slip a week or two, but I would personally push for re-landing it then "regardless" of the regressions. We should try to focus efforts on understanding and removing as much of the performance impact as possible, as everyone pretty much agrees that it should be quite minimal (and entirely in the linker, basically).
r? @nnethercote
Stabilize control-flow-guard codegen option
This is the stabilization PR discussed in #68793. It converts the `-Z control-flow-guard` debugging option into a codegen option (`-C control-flow-guard`), and changes the associated tests.
Teach bootstrap install and dist commands about TargetSelection
With this, we can now use a target JSON file to build a
cross-compiler:
```
x.py install --host ../aarch64-apple-darwin.json --target aarch64-apple-darwin
```
r? @Mark-Simulacrum
With this, we can now use a target JSON file to build a
cross-compiler:
```
x.py install --host ../aarch64-apple-darwin.json --target aarch64-apple-darwin
```
Generating the coverage map
@tmandry @wesleywiser
rustc now generates the coverage map and can support (limited)
coverage report generation, at the function level.
Example commands to generate a coverage report:
```shell
$ BUILD=$HOME/rust/build/x86_64-unknown-linux-gnu
$ $BUILD/stage1/bin/rustc -Zinstrument-coverage \
$HOME/rust/src/test/run-make-fulldeps/instrument-coverage/main.rs
$ LLVM_PROFILE_FILE="main.profraw" ./main
called
$ $BUILD/llvm/bin/llvm-profdata merge -sparse main.profraw -o main.profdata
$ $BUILD/llvm/bin/llvm-cov show --instr-profile=main.profdata main
```
r? @wesleywiser
Rust compiler MCP rust-lang/compiler-team#278
Relevant issue: #34701 - Implement support for LLVMs code coverage instrumentation
bootstrap.py: patch RPATH on NixOS to handle the new zlib dependency.
This is a stop-gap until #74420 is resolved (assuming we'll patch beta to statically link zlib).
However, I've been meaning to rewrite the NixOS support we have in `bootstrap.py` for a while now, and had to in order to cleanly add zlib as a dependency (the second commit is a relatively small delta in functionality, compared to the first).
Previously, we would extract the `ld-linux.so` path from the output of `ldd /run/current-system/sw/bin/sh`, which assumes a lot. On top of that we didn't use any symlinks, which meant if the user ran GC (`nix-collect-garbage`), e.g. after updating their system, their `stage0` binaries would suddenly be broken (i.e. referring to files that no longer exist).
We were also using `patchelf` directly, assuming it can be found in `$PATH` (which is not necessarily true).
My new approach relies on using `nix-build` to get the following "derivations" (packages, more or less):
* `stdenv.cc.bintools`, which has a `nix-support/dynamic-linker` file containing the path to `ld-linux.so`
* reading this file is [the canonical way to run `patchelf --set-interpreter`](https://github.com/NixOS/nixpkgs/search?l=Nix&q=%22--set-interpreter+%24%28cat+%24NIX_CC%2Fnix-support%2Fdynamic-linker%29%22)
* `patchelf` (so that the user doesn't need to have it installed)
* `zlib`, for the `libz.so` dependency of `libLLVM-*.so` (until #74420 is resolved, presumably)
This is closer to how software is built on Nix, but I've tried to keep it as simple as possible (and not add e.g. a `stage0.nix` file).
Symlinks to each of those dependencies are kept in `stage0/.nix-deps`, which prevents GC from invalidating `stage0` binaries.
r? @nagisa cc @Mark-Simulacrum @oli-obk @davidtwco
Teach bootstrap about target files vs target triples
`rustc` allows passing in predefined target triples as well as JSON
target specification files. This change allows bootstrap to have the
first inkling about those differences. This allows building a
cross-compiler for an out-of-tree architecture (even though that
compiler won't work for other reasons).
Even if no one ever uses this functionality, I think the newtype
around the `Interned<String>` improves the readability of the code.
Enforce even more the code blocks attributes check through rustdoc
`rustdoc` now has a lint which allows it to warn if a code block attribute is malformated (which can end up in bad situations, even more in case of testing examples!). Now it'll fail if such a situation is encountered when testing markdown code blocks examples.
r? @Mark-Simulacrum
`rustc` allows passing in predefined target triples as well as JSON
target specification files. This change allows bootstrap to have the
first inkling about those differences. This allows building a
cross-compiler for an out-of-tree architecture (even though that
compiler won't work for other reasons).
Even if no one ever uses this functionality, I think the newtype
around the `Interned<String>` improves the readability of the code.
