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.
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.
The current plan is that submodule tracks the `release` branch of
rust-analyzer, which is updated once a week.
rust-analyzer is a workspace (with a virtual manifest), the actual
binary is provide by `crates/rust-analyzer` package.
Note that we intentionally don't add rust-analyzer to `Kind::Test`,
for two reasons.
*First*, at the moment rust-analyzer's test suite does a couple of
things which might not work in the context of rust repository. For
example, it shells out directly to `rustup` and `rustfmt`. So, making
this work requires non-trivial efforts.
*Second*, it seems unlikely that running tests in rust-lang/rust repo
would provide any additional guarantees. rust-analyzer builds with
stable and does not depend on the specifics of the compiler, so
changes to compiler can't break ra, unless they break stability
guarantee. Additionally, rust-analyzer itself is gated on bors, so we
are pretty confident that test suite passes.
With #65251 landed there's no need to build two LLVM backends and ship
them with rustc, every target we have now uses the same LLVM backend!
This removes the `src/llvm-emscripten` submodule and additionally
removes all support from rustbuild for building the emscripten LLVM
backend. Multiple codegen backend support is left in place for now, and
this is intended to be an easy 10-15 minute win on CI times by avoiding
having to build LLVM twice.
Update to LLVM 9 trunk
Following the preparatory changes in #62474, this updates the LLVM submodule to https://github.com/rust-lang/llvm-project/tree/rustc/9.0-2019-07-12 and:
* Changes the LLVM Rust bindings to account for the new SubtargetSubTypeKV.
* Adjusts a codegen test for the new form of the byval attribute that takes a type.
* Makes a PGO codegen test more liberal with regard to order and linkage.
* Builds InstrProfilingPlatformWindows.c as part of libprofiler_builtins.
* Moves registration of additional passes (in particular sanitizers) to the end of the module pass manager.
* Disables LLDB on builders.
r? @alexcrichton
The new git submodule src/llvm-project is a monorepo replacing src/llvm
and src/tools/{clang,lld,lldb}. This also serves as a rebase for these
projects to the new 8.x branch from trunk.
The src/llvm-emscripten fork is unchanged for now.
This commit switches the standard library to using the `backtrace-sys`
crate from crates.io instead of duplicating the logic here in the Rust
repositor with the `backtrace-sys`'s crate's logic.
Eventually this will hopefully be a good step towards using the
`backtrace` crate directly from crates.io itself, but we're not quite
there yet! Hopefully this is a small incremental first step we can take.
Ever since we added a Cargo-based build system for the compiler the
standard library has always been a little special, it's never been able
to depend on crates.io crates for runtime dependencies. This has been a
result of various limitations, namely that Cargo doesn't understand that
crates from crates.io depend on libcore, so Cargo tries to build crates
before libcore is finished.
I had an idea this afternoon, however, which lifts the strategy
from #52919 to directly depend on crates.io crates from the standard
library. After all is said and done this removes a whopping three
submodules that we need to manage!
The basic idea here is that for any crate `std` depends on it adds an
*optional* dependency on an empty crate on crates.io, in this case named
`rustc-std-workspace-core`. This crate is overridden via `[patch]` in
this repository to point to a local crate we write, and *that* has a
`path` dependency on libcore.
Note that all `no_std` crates also depend on `compiler_builtins`, but if
we're not using submodules we can publish `compiler_builtins` to
crates.io and all crates can depend on it anyway! The basic strategy
then looks like:
* The standard library (or some transitive dep) decides to depend on a
crate `foo`.
* The standard library adds
```toml
[dependencies]
foo = { version = "0.1", features = ['rustc-dep-of-std'] }
```
* The crate `foo` has an optional dependency on `rustc-std-workspace-core`
* The crate `foo` has an optional dependency on `compiler_builtins`
* The crate `foo` has a feature `rustc-dep-of-std` which activates these
crates and any other necessary infrastructure in the crate.
A sample commit for `dlmalloc` [turns out to be quite simple][commit].
After that all `no_std` crates should largely build "as is" and still be
publishable on crates.io! Notably they should be able to continue to use
stable Rust if necessary, since the `rename-dependency` feature of Cargo
is soon stabilizing.
As a proof of concept, this commit removes the `dlmalloc`,
`libcompiler_builtins`, and `libc` submodules from this repository. Long
thorns in our side these are now gone for good and we can directly
depend on crates.io! It's hoped that in the long term we can bring in
other crates as necessary, but for now this is largely intended to
simply make it easier to manage these crates and remove submodules.
This should be a transparent non-breaking change for all users, but one
possible stickler is that this almost for sure breaks out-of-tree
`std`-building tools like `xargo` and `cargo-xbuild`. I think it should
be relatively easy to get them working, however, as all that's needed is
an entry in the `[patch]` section used to build the standard library.
Hopefully we can work with these tools to solve this problem!
[commit]: 28ee12db81
This commit removes all jemalloc related submodules, configuration, etc,
from the bootstrap, from the standard library, and from the compiler.
This will be followed up with a change to use jemalloc specifically as
part of rustc on blessed platforms.
Fixes a problem where submodules could not be cloned under some git
configurations. Specifically, when url.git@github.com:.insteadOf =
https://github.com/ is set.
commit 6c10142251 ("Update LLVM submodule") disabled the lldb build.
This patch updates the lldb and clang submodules to once again build
against the LLVM that is included in the Rust tree, and reverts the
.travis.yml changes from that patch.
This optionally adds lldb (and clang, which it needs) to the build.
Because rust uses LLVM 7, and because clang 7 is not yet released, a
recent git master version of clang is used.
The lldb that is used includes the Rust plugin.
lldb is only built when asked for, or when doing a nightly build on
macOS. Only macOS is done for now due to difficulties with the Python
dependency.
While we're at it update the `backtrace` crate from crates.io. It turns out that
the submodule's configure script has gotten a lot more finnicky as of late so
also switch over to using the `cc` crate manually which allows to avoid some
hacks around the configure script as well
This commit imports the LLD project from LLVM to serve as the default linker for
the `wasm32-unknown-unknown` target. The `binaryen` submoule is consequently
removed along with "binaryen linker" support in rustc.
Moving to LLD brings with it a number of benefits for wasm code:
* LLD is itself an actual linker, so there's no need to compile all wasm code
with LTO any more. As a result builds should be *much* speedier as LTO is no
longer forcibly enabled for all builds of the wasm target.
* LLD is quickly becoming an "official solution" for linking wasm code together.
This, I believe at least, is intended to be the main supported linker for
native code and wasm moving forward. Picking up support early on should help
ensure that we can help LLD identify bugs and otherwise prove that it works
great for all our use cases!
* Improvements to the wasm toolchain are currently primarily focused around LLVM
and LLD (from what I can tell at least), so it's in general much better to be
on this bandwagon for bugfixes and new features.
* Historical "hacks" like `wasm-gc` will soon no longer be necessary, LLD
will [natively implement][gc] `--gc-sections` (better than `wasm-gc`!) which
means a postprocessor is no longer needed to show off Rust's "small wasm
binary size".
LLD is added in a pretty standard way to rustc right now. A new rustbuild target
was defined for building LLD, and this is executed when a compiler's sysroot is
being assembled. LLD is compiled against the LLVM that we've got in tree, which
means we're currently on the `release_60` branch, but this may get upgraded in
the near future!
LLD is placed into rustc's sysroot in a `bin` directory. This is similar to
where `gcc.exe` can be found on Windows. This directory is automatically added
to `PATH` whenever rustc executes the linker, allowing us to define a `WasmLd`
linker which implements the interface that `wasm-ld`, LLD's frontend, expects.
Like Emscripten the LLD target is currently only enabled for Tier 1 platforms,
notably OSX/Windows/Linux, and will need to be installed manually for compiling
to wasm on other platforms. LLD is by default turned off in rustbuild, and
requires a `config.toml` option to be enabled to turn it on.
Finally the unstable `#![wasm_import_memory]` attribute was also removed as LLD
has a native option for controlling this.
[gc]: https://reviews.llvm.org/D42511
This commit imports the `stdsimd` crate into the standard library,
creating an `arch` and `simd` module inside of both libcore and libstd.
Both of these modules are **unstable** and will continue to be so until
RFC 2335 is stabilized.
As a brief recap, the modules are organized as so:
* `arch` contains all current architectures with intrinsics, for example
`std::arch::x86`, `std::arch::x86_64`, `std::arch::arm`, etc. These
modules contain all of the intrinsics defined for the platform, like
`_mm_set1_epi8`.
* In the standard library, the `arch` module also exports a
`is_target_feature_detected` macro which performs runtime detection to
determine whether a target feature is available at runtime.
* The `simd` module contains experimental versions of strongly-typed
lane-aware SIMD primitives, to be fully fleshed out in a future RFC.
The main purpose of this commit is to start pulling in all these
intrinsics and such into the standard library on nightly and allow
testing and such. This'll help allow users to easily kick the tires and
see if intrinsics work as well as allow us to test out all the
infrastructure for moving the intrinsics into the standard library.
Is it really time? Have our months, no, *years* of suffering come to an end? Are we finally able to cast off the pall of Hoedown? The weight which has dragged us down for so long?
-----
So, timeline for those who need to catch up:
* Way back in December 2016, [we decided we wanted to switch out the markdown renderer](https://github.com/rust-lang/rust/issues/38400). However, this was put on hold because the build system at the time made it difficult to pull in dependencies from crates.io.
* A few months later, in March 2017, [the first PR was done, to switch out the renderers entirely](https://github.com/rust-lang/rust/pull/40338). The PR itself was fraught with CI and build system issues, but eventually landed.
* However, not all was well in the Rustdoc world. During the PR and shortly after, we noticed [some differences in the way the two parsers handled some things](https://github.com/rust-lang/rust/issues/40912), and some of these differences were major enough to break the docs for some crates.
* A couple weeks afterward, [Hoedown was put back in](https://github.com/rust-lang/rust/pull/41290), at this point just to catch tests that Pulldown was "spuriously" running. This would at least provide some warning about spurious tests, rather than just breaking spontaneously.
* However, the problems had created enough noise by this point that just a few days after that, [Hoedown was switched back to the default](https://github.com/rust-lang/rust/pull/41431) while we came up with a solution for properly warning about the differences.
* That solution came a few weeks later, [as a series of warnings when the HTML emitted by the two parsers was semantically different](https://github.com/rust-lang/rust/pull/41991). But that came at a cost, as now rustdoc needed proc-macro support (the new crate needed some custom derives farther down its dependency tree), and the build system was not equipped to handle it at the time. It was worked on for three months as the issue stumped more and more people.
* In that time, [bootstrap was completely reworked](https://github.com/rust-lang/rust/pull/43059) to change how it ordered compilation, and [the method by which it built rustdoc would change](https://github.com/rust-lang/rust/pull/43482), as well. This allowed it to only be built after stage1, when proc-macros would be available, allowing the "rendering differences" PR to finally land.
* The warnings were not perfect, and revealed a few [spurious](https://github.com/rust-lang/rust/pull/44368) [differences](https://github.com/rust-lang/rust/pull/45421) between how we handled the renderers.
* Once these were handled, [we flipped the switch to turn on the "rendering difference" warnings all the time](https://github.com/rust-lang/rust/pull/45324), in October 2017. This began the "warning cycle" for this change, and landed in stable in 1.23, on 2018-01-04.
* Once those warnings hit stable, and after a couple weeks of seeing whether we would get any more reports than what we got from sitting on nightly/beta, [we switched the renderers](https://github.com/rust-lang/rust/pull/47398), making Pulldown the default but still offering the option to use Hoedown.
And that brings us to the present. We haven't received more new issues from this in the meantime, and the "switch by default" is now on beta. Our reasoning is that, at this point, anyone who would have been affected by this has run into it already.
This commit introduces a separately compiled backend for Emscripten, avoiding
compiling the `JSBackend` target in the main LLVM codegen backend. This builds
on the foundation provided by #47671 to create a new codegen backend dedicated
solely to Emscripten, removing the `JSBackend` of the main codegen backend in
the process.
A new field was added to each target for this commit which specifies the backend
to use for translation, the default being `llvm` which is the main backend that
we use. The Emscripten targets specify an `emscripten` backend instead of the
main `llvm` one.
There's a whole bunch of consequences of this change, but I'll try to enumerate
them here:
* A *second* LLVM submodule was added in this commit. The main LLVM submodule
will soon start to drift from the Emscripten submodule, but currently they're
both at the same revision.
* Logic was added to rustbuild to *not* build the Emscripten backend by default.
This is gated behind a `--enable-emscripten` flag to the configure script. By
default users should neither check out the emscripten submodule nor compile
it.
* The `init_repo.sh` script was updated to fetch the Emscripten submodule from
GitHub the same way we do the main LLVM submodule (a tarball fetch).
* The Emscripten backend, turned off by default, is still turned on for a number
of targets on CI. We'll only be shipping an Emscripten backend with Tier 1
platforms, though. All cross-compiled platforms will not be receiving an
Emscripten backend yet.
This commit means that when you download the `rustc` package in Rustup for Tier
1 platforms you'll be receiving two trans backends, one for Emscripten and one
that's the general LLVM backend. If you never compile for Emscripten you'll
never use the Emscripten backend, so we may update this one day to only download
the Emscripten backend when you add the Emscripten target. For now though it's
just an extra 10MB gzip'd.
Closes#46819
This commit adds a new target to the compiler: wasm32-unknown-unknown. This
target is a reimagining of what it looks like to generate WebAssembly code from
Rust. Instead of using Emscripten which can bring with it a weighty runtime this
instead is a target which uses only the LLVM backend for WebAssembly and a
"custom linker" for now which will hopefully one day be direct calls to lld.
Notable features of this target include:
* There is zero runtime footprint. The target assumes nothing exists other than
the wasm32 instruction set.
* There is zero toolchain footprint beyond adding the target. No custom linker
is needed, rustc contains everything.
* Very small wasm modules can be generated directly from Rust code using this
target.
* Most of the standard library is stubbed out to return an error, but anything
related to allocation works (aka `HashMap`, `Vec`, etc).
* Naturally, any `#[no_std]` crate should be 100% compatible with this new
target.
This target is currently somewhat janky due to how linking works. The "linking"
is currently unconditional whole program LTO (aka LLVM is being used as a
linker). Naturally that means compiling programs is pretty slow! Eventually
though this target should have a linker.
This target is also intended to be quite experimental. I'm hoping that this can
act as a catalyst for further experimentation in Rust with WebAssembly. Breaking
changes are very likely to land to this target, so it's not recommended to rely
on it in any critical capacity yet. We'll let you know when it's "production
ready".
---
Currently testing-wise this target is looking pretty good but isn't complete.
I've got almost the entire `run-pass` test suite working with this target (lots
of tests ignored, but many passing as well). The `core` test suite is still
getting LLVM bugs fixed to get that working and will take some time. Relatively
simple programs all seem to work though!
---
It's worth nothing that you may not immediately see the "smallest possible wasm
module" for the input you feed to rustc. For various reasons it's very difficult
to get rid of the final "bloat" in vanilla rustc (again, a real linker should
fix all this). For now what you'll have to do is:
cargo install --git https://github.com/alexcrichton/wasm-gc
wasm-gc foo.wasm bar.wasm
And then `bar.wasm` should be the smallest we can get it!
---
In any case for now I'd love feedback on this, particularly on the various
integration points if you've got better ideas of how to approach them!
