[clang][WebAssembly] Support wasm32-wasi shared libraries #148
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This adds support for Emscripten-style shared libraries [1] to non-emscripten targets, such as `wasm32-wasi`. Previously, only static linking was supported, and the `-shared` and `-fPIC` flags were simply ignored. Now both flags are honored. Since WASI runtimes do not necessarily include JavaScript support, we cannot rely on the JS-based Emscripten linker to link shared libraries. Instead, we link them using the Component Model proposal [2]. We have prototyped shared library support in `wasi-sdk` [3] and put together a demo [4] which uses a patched version of `wit-component` [5] to link libraries using the Component Model. We plan to submit the required changes upstream to the respective repositories in the next week or two. [1] https://github.com/WebAssembly/tool-conventions/blob/main/DynamicLinking.md [2] https://github.com/WebAssembly/component-model/blob/main/design/mvp/examples/SharedEverythingDynamicLinking.md [3] https://github.com/dicej/wasi-sdk/tree/dynamic-linking [4] https://github.com/dicej/component-linking-demo [5] https://github.com/bytecodealliance/wasm-tools/tree/main/crates/wit-component Signed-off-by: Joel Dice <[email protected]> Reviewed By: sbc100 Differential Revision: https://reviews.llvm.org/D153293 Signed-off-by: Joel Dice <[email protected]>
|
What will it look like to be used in the rust compiler? |
|
This patch alone won't change the behavior of the rust compiler itself, but it will allow you to build the standard library and application code with Eventually, we'll want to make that whole process easier, but this at least makes it possible. I'm also working with Dan Gohman on changes to |
|
OK, as long as we're not changing that |
bors
pushed a commit
to rust-lang-ci/rust
that referenced
this pull request
Jul 27, 2023
bors
added a commit
to rust-lang-ci/rust
that referenced
this pull request
Jul 27, 2023
Update LLVM submodule This adds rust-lang/llvm-project#148.
trvswgnr
added a commit
to crablang/crab
that referenced
this pull request
Jul 28, 2023
* Dynamic for smir * Optimize format usage Per #112156, using `&` in `format!` may cause a small perf delay, so I tried to clean up one module at a time format usage. This PR includes a few removals of the ref in format (they do compile locally without the ref), as well as a few format inlining for consistency. * compiler: Hermit targets: Remove pre-link args. These pre-link args are remains from Hermit's old C version. We don't need them and we have no reason to override the defaults here. See ld [1] for details. [1]: https://sourceware.org/binutils/docs/ld/Options.html Signed-off-by: Martin Kröning <[email protected]> * compiler: Hermit targets: Sort base fields by declaration Signed-off-by: Martin Kröning <[email protected]> * compiler: Hermit targets: Use functional update syntax instead of mutating the base. Signed-off-by: Martin Kröning <[email protected]> * compiler: Add `riscv64gc-unknown-hermit` target Co-authored-by: Martin Kröning <[email protected]> Signed-off-by: Martin Kröning <[email protected]> * compiler: Add `*-unknown-hermit` documentation Signed-off-by: Martin Kröning <[email protected]> * Remove redundant note. This came from x86_64-unknown-none and doesn't make sense here. * Add clarification about build-std and using newer instructions. * new unstable option: -Zwrite-long-types-to-disk This option guards the logic of writing long type names in files and instead using short forms in error messages in rustc_middle/ty/error behind a flag. The main motivation for this change is to disable this behaviour when running ui tests. This logic can be triggered by running tests in a directory that has a long enough path, e.g. /my/very-long-path/where/rust-codebase/exists/ This means ui tests can fail depending on how long the path to their file is. Some ui tests actually rely on this behaviour for their assertions, so for those we enable the flag manually. * coverage: Obtain the `__llvm_covfun` section name outside a per-function loop This section name is always constant for a given target, but obtaining it from LLVM requires a few intermediate allocations. There's no need to do so repeatedly from inside a per-function loop. * interpret: support projecting into Place::Local without force_allocation * Explain what the heck is going on with this lifetime remapping business * builtin_macros: expect raw strings too `expr_to_string` allows raw strings through so this code should be expected to handle those. Signed-off-by: David Wood <[email protected]> * Some documentation nits * Fix missing attribute merge on glob foreign re-exports * Add regression test for #113982 * Hide `ToString` implementations that specialize the default ones The status quo is highly confusing, since the overlap is not apparent, and specialization is not a feature of Rust. This addresses #87545; I'm not certain if it closes it, since that issue might also be trackign a *general* solution for hiding specializing impls automatically. * borrowck/errors: fix i18n error in delayed bug During borrowck, the `MultiSpan` from a buffered diagnostic is cloned and used to emit a delayed bug indicating a diagnostic was buffered - when the buffered diagnostic is translated, then the cloned `MultiSpan` may contain labels which can only render with the diagnostic's arguments, but the delayed bug being emitted won't have those arguments. Adds a function which clones `MultiSpan` without also cloning the contained labels, and use this function when creating the buffered diagnostic delayed bug. Signed-off-by: David Wood <[email protected]> * Add missing documentation for `Session::time` * If re-export is private, get the next item until a public one is found or expose the private item directly * Add regression test for #81141 * Improve code readability * Extend issue-81141-private-reexport-in-public-api test to cover more cases * Rename `first_not_private` into `first_non_private` * Correctly handle `super` and `::` * Add test for private items * Add support for `--document-hidden-items` in `first_non_private` * Add test for `--document-hidden-items` * Re-add missing generics in `first_non_private` * Add regression test for generics reexport of private import * Remove needs for transmute * Revert "Remove needs for transmute" This reverts commit ea9a17b9995b7a076283777b7d462a360fece2d6. * Cache qpath first public result * Fix span for punnycode * Explain RPITs in the way they actually work * lcnr's suggestions Co-authored-by: lcnr <[email protected]> * validate `doc(masked)` * compiler: Add `x86_64-unikraft-linux-musl` target Signed-off-by: Martin Kröning <[email protected]> * compiler: Add `*-unikraft-linux-musl` documentation Signed-off-by: Martin Kröning <[email protected]> * bootstrap: Don't bundle musl on Unikraft Signed-off-by: Martin Kröning <[email protected]> * `unix::init`: Handle `ENOSYS` from `poll` on Unikraft. Signed-off-by: Martin Kröning <[email protected]> * `unix::init`: Don't use `signal` on Unikraft. Signed-off-by: Martin Kröning <[email protected]> * Update cargo * fix(resolve): skip panic when resolution is dummy * resolve: ensure compile failed when has dummy or ambiguous * typos * delete [allow(...)] from issue #74838 * remove additional [allow(unused_unsafe)] * bootstrap: Define CMake platform if DragonFly. CMAKE_SYSTEM_NAME is defined on a cross build if the target is recognized. Without this explicit definition cmake will assume that we're building for the host platform which can bring in unwanted compiler and linker flags. Also, add a warning on cross builds with unknown target to aid in cross builds for future platforms. * docs(LazyLock): add example pass local LazyLock variable to struct Signed-off-by: DragonBillow <[email protected]> * factor out more stable impls * Added recursive checking back up to see if a `Clone` suggestion would be helpful. * extract common code * builtin_macros: raw str in diagnostic output If a raw string was used in the `env!` invocation, then it should also be shown in the diagnostic messages as a raw string. Signed-off-by: David Wood <[email protected]> * Add regression test for invalid unused const in method The warning can be reproduced with 1.63 but not with 1.64. $ rustc +1.63 tests/ui/lint/unused/const-local-var.rs warning: constant `F` is never used --> tests/ui/lint/unused/const-local-var.rs:14:9 | 14 | const F: i32 = 2; | ^^^^^^^^^^^^^^^^^ | = note: `#[warn(dead_code)]` on by default $ rustc +1.64 tests/ui/lint/unused/const-local-var.rs Add a regression test to prevent the problem from re-appearing. * clippy: `env!` invocations can't be b"" literals Signed-off-by: David Wood <[email protected]> * write-long-types-to-disk: update tests * interpret: refactor projection code to work on a common trait, and use that for visitors * interpret: read_discriminant: only return VariantIdx * bless more * add some sanity checks in write_immediate_no_validate * Suggest `{Option,Result}::as_ref()` instead of `cloned()` in some cases * Remove some arguments that are always the same * Inline a function that is only used once * Remove a redundant field * Improve performance of `first_non_private` * Try explaining where `Inner` is in the signature better * Use a builder instead of boolean/option arguments * abi: unsized field in union - assert to delay bug Unions cannot have unsized fields, and as such, layout computation for unions asserts that each union field is sized (as this would normally have halted compilation earlier). However, if a generator ends up with an unsized local - a circumstance in which an error will always have been emitted earlier, for example, if attempting to dereference a `&str` - then the generator transform will produce a union with an unsized field. Since #110107, later passes will be run, such as constant propagation, and can attempt layout computation on the generator, which will result in layout computation of `str` in the context of it being a field of a union - and so the aforementioned assertion would cause an ICE. It didn't seem appropriate to try and detect this case in the MIR body and skip this specific pass; tainting the MIR body or delaying a bug from the generator transform (or elsewhere) wouldn't prevent this either (as neither would prevent the later pass from running); and tainting when the deref of `&str` is reported, if that's possible, would unnecessarily prevent potential other errors from being reported later in compilation, and is very tailored to this specific case of getting a unsized type in a generator. Given that this circumstance can only happen when an error should have already been reported, the correct fix appears to be just changing the assert to a delayed bug. This will still assert if there is some circumstance where this occurs and no error has been reported, but it won't crash the compiler in this instance. Signed-off-by: David Wood <[email protected]> * Normalize the RHS of an unsize goal * Make sure to detect trait upcasting coercion even after normalization * Restore tuple unsizing feature gate * Consolidate trait upcasting and unsize into one normalization * CI: fix CMake installation for 32 and 64bit `dist` Linux * Add help for crate arg when crate name is invalid * Make everything builtin! * Remove credential providers from bootstrap since they are now built-in to the Cargo binary * Split nested GHA groups instead of panicking * interpret: make write functions generic over the place type * interpret: make read functions generic over operand type * make MPlaceTy non-Copy * Remove -Z diagnostic-width * inline format!() args from rustc_codegen_llvm to the end (4) r? @WaffleLapkin * Optimize `AtomicBool` for target that don't support byte-sized atomics `AtomicBool` is defined to have the same layout as `bool`, which means that we guarantee that it has a size of 1 byte. However on certain architectures such as RISC-V, LLVM will emulate byte atomics using a masked CAS loop on an aligned word. We can take advantage of the fact that `bool` only ever has a value of 0 or 1 to replace `swap` operations with `and`/`or` operations that LLVM can lower to word-sized atomic `and`/`or` operations. This takes advantage of the fact that the incoming value to a `swap` or `compare_exchange` for `AtomicBool` is often a compile-time constant. * Remove `Parser::desugar_doc_comments`. It's currently stored twice: once in `Parser`, once in the `TokenStream` within `Parser`. We only need the latter. * Add `sym::iter_mut` + `sym::as_mut_ptr` * rustdoc: stylistic changes * rustdoc: fix cross-crate impl-Sized * Don't treat negative trait predicates as always knowable * Tweak `Parser::look_ahead`. It doesn't really matter what the `desugar_doc_comments` argument is here, because in practice we never look ahead through doc comments. Changing it to `cursor.desugar_doc_comments` will allow some follow-up simplifications. * Remove `desugar_doc_comments` arguments from `TokenCursor::{inlined_,}next`. Because it's now always `self.desugar_doc_comments`. * Add a comment to `TokenCursor::desugar_doc_comments`. Useful information that took me some time to discern. * add tidy check that forbids issue ui test filenames * add stable NullaryOp * Make `x test src/tools/rustfmt --bless` format rustfmt with the freshly built in-tree version * Bump syn now that it doesn't affect diagnostics anymore * Bump syn dependency * valtree: a bit of cleanup * compiletest: remove ci-specific remap-path-prefix Now that we have fixed the underlying cause of long type name inconsistencies in #113893, we can remove the remap-path-prefix logic from CI * Regression test `println!()` panic message on `ErrorKind::BrokenPipe` No existing test failed if the [`panic!()`][1] of the `println!()` family of functions was removed, or if its message was changed. So add such a test. [1] https://github.com/rust-lang/rust/blob/104f4300cfddbd956e32820ef202a732f06ec848/library/std/src/io/stdio.rs#L1007-L1009 * docs: fmt::Debug*: Fix comments for finish method. In the code sample for the `finish` method on `DebugList`, `DebugMap`, and `DebugSet`, refer to finishing the list, map, or set, rather than struct as it did. * Add Param ty to SMIR * Add Bound ty to SMIR * Have a better file name than just the issue id * Fix regression for private in public * Replace in-tree `rustc_apfloat` with the new version of the crate * Fix miri * Remove unused NCSA license * replace atty crate with std's isTerminal * Squelch a noisy rustc_expand unittest * Add test case for #109567 * Add tests for #102403 and #113407 * Dont pass -Zwrite-long-types-to-disk=no for ui-fulldeps --stage=1 * Unite bless environment variables under `RUSTC_BLESS` Currently, Clippy, Miri, Rustfmt, and rustc all use an environment variable to indicate that output should be blessed, but they use different variable names. In order to improve consistency, this patch applies the following changes: - Emit `RUSTC_BLESS` within `prepare_cargo_test` so it is always available - Change usage of `MIRI_BLESS` in the Miri subtree to use `RUSTC_BLESS` - Change usage of `BLESS` in the Clippy subtree to `RUSTC_BLESS` - Change usage of `BLESS` in the Rustfmt subtree to `RUSTC_BLESS` - Adjust the blessable test in `rustc_errors` to use this same convention - Update documentation where applicable Any tools that uses `RUSTC_BLESS` should check that it is set to any value other than `"0"`. * Update LLVM submodule This adds rust-lang/llvm-project#148. * Fix URL for `rmatches` * Simplify the `ttdelim_span` test. The existing code is a very complex and inefficient way to the get the span of the last token. * Replace `into_trees` with `trees` in a test. There's no need for token tree cloning here. * Use `TokenStream::trees` instead of `into_trees` for attributes. This avoids cloning some token trees. A couple of `clone` calls were inserted, but only on some paths, and the next commit will remove them. * Make `TokenTree::uninterpolate` take `&self` and return a `Cow`. Making it similar to `Token::uninterpolate`. This avoids some more token tree cloning. * Avoid some token cloning in `filter_tokens_from_list`. Now the cloning only happens on some paths, instead of all paths. * Avoid some token tree cloning in decl macro parsing. By changing `into_trees` into `trees`. Some of the subsequent paths require explicit clones, but not all. * Avoid `into_trees` usage in rustfmt. Token tree cloning is only needed in one place. * Remove `Iterator` impl for `TokenTreeCursor`. This is surprising, but the new comment explains why. It's a logical conclusion in the drive to avoid `TokenTree` clones. `TokenTreeCursor` is now only used within `Parser`. It's still needed due to `replace_prev_and_rewind`. * Revert "don't uniquify regions when canonicalizing" This reverts commit 171f541. * Consider a goal as NOT changed if its response is identity modulo regions * Optimize `TokenKind::clone`. `TokenKind` would impl `Copy` if it weren't for `TokenKind::Interpolated`. This commit makes `clone` reflect that. * tests/ui/hello_world/main.rs: Remove FIXME The purpose of the test is to make sure that compiling hello world produces no compiler output. To properly test that, we need to run the entire compiler pipeline. We don't want the test to pass if codegen accidentally starts writing to stdout. So keep it as build-pass. * Turns out opaque types can have hidden types registered during mir validation * Remove transmute calls and caching for use paths * Revert "add tidy check that forbids issue ui test filenames" This reverts commit 13e2abf. Reverting because an MCP was requested and it turned out there was a lack of a consensus on what to do in this area. * Remove `constness` from `ParamEnv` * update tests, adding known-bug * Make `--print KIND=PATH` unstable rust-lang/rust#113780 should have gone through an MCP+FCP but wasn't, but instead of reverting the original PR, this PR just make that new option unstable. * bless clippy * docs(style-guide): don't flatten match arms with macro call * Change the description of `SUSPICIOUS_AUTO_TRAIT_IMPLS` * Introduce the `#[diagnostic]` attribute namespace Co-authored-by: est31 <[email protected]> Co-authored-by: Esteban Kuber <[email protected]> Co-authored-by: Vadim Petrochenkov <[email protected]> * imported src/doc/nomicon into main repository * imported src/tools/cargo into main repository * imported src/doc/reference into main repository * imported src/doc/book into main repository * imported src/doc/rust-by-example into main repository * imported library/stdarch into main repository * imported src/doc/rustc-dev-guide into main repository * imported src/doc/edition-guide into main repository * imported src/doc/embedded-book into main repository * imported library/backtrace into main repository * use crablang readme * imported src/tools/cargo into main repository * abandon submodules added a bash script that converts a project using gitsubmodules (ew) to a monorepo * use crab contrib --------- Signed-off-by: Martin Kröning <[email protected]> Signed-off-by: David Wood <[email protected]> Signed-off-by: DragonBillow <[email protected]> Co-authored-by: Eric Mark Martin <[email protected]> Co-authored-by: Yuri Astrakhan <[email protected]> Co-authored-by: bors <[email protected]> Co-authored-by: Martin Kröning <[email protected]> Co-authored-by: Simon Schöning <[email protected]> Co-authored-by: Laurențiu Nicola <[email protected]> Co-authored-by: Jonathan Pallant (Ferrous Systems) <[email protected]> Co-authored-by: Mahdi Dibaiee <[email protected]> Co-authored-by: Zalathar <[email protected]> Co-authored-by: Ralf Jung <[email protected]> Co-authored-by: Oli Scherer <[email protected]> Co-authored-by: David Wood <[email protected]> Co-authored-by: Guillaume Gomez <[email protected]> Co-authored-by: Frank Steffahn <[email protected]> Co-authored-by: Eric Huss <[email protected]> Co-authored-by: Michael Goulet <[email protected]> Co-authored-by: lcnr <[email protected]> Co-authored-by: Matthias Krüger <[email protected]> Co-authored-by: Lukas Markeffsky <@> Co-authored-by: Martin Kröning <[email protected]> Co-authored-by: Arlo Siemsen <[email protected]> Co-authored-by: bohan <[email protected]> Co-authored-by: Tshepang Mbambo <[email protected]> Co-authored-by: James Dietz <[email protected]> Co-authored-by: Alex Zepeda <[email protected]> Co-authored-by: DragonBillow <[email protected]> Co-authored-by: Steven Trotter <[email protected]> Co-authored-by: Deadbeef <[email protected]> Co-authored-by: Martin Nordholts <[email protected]> Co-authored-by: clubby789 <[email protected]> Co-authored-by: Jakub Beránek <[email protected]> Co-authored-by: yukang <[email protected]> Co-authored-by: Amanieu d'Antras <[email protected]> Co-authored-by: Nicholas Nethercote <[email protected]> Co-authored-by: blyxyas <[email protected]> Co-authored-by: León Orell Valerian Liehr <[email protected]> Co-authored-by: asquared31415 <[email protected]> Co-authored-by: Bruce Mitchener <[email protected]> Co-authored-by: Santiago Pastorino <[email protected]> Co-authored-by: MoskalykA <[email protected]> Co-authored-by: Wesley Wiser <[email protected]> Co-authored-by: klensy <[email protected]> Co-authored-by: Trevor Gross <[email protected]> Co-authored-by: Joel Dice <[email protected]> Co-authored-by: Veera <[email protected]> Co-authored-by: Guillaume Gomez <[email protected]> Co-authored-by: Jubilee Young <[email protected]> Co-authored-by: Urgau <[email protected]> Co-authored-by: Jubilee <[email protected]> Co-authored-by: Caleb Cartwright <[email protected]> Co-authored-by: Oleksandr Babak <[email protected]> Co-authored-by: Georg Semmler <[email protected]> Co-authored-by: Vadim Petrochenkov <[email protected]>
github-actions bot
pushed a commit
to rust-lang/miri
that referenced
this pull request
Jul 30, 2023
Update LLVM submodule This adds rust-lang/llvm-project#148.
vext01
pushed a commit
to vext01/llvm-project
that referenced
this pull request
May 1, 2024
Serialise PHI nodes.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This adds support for Emscripten-style shared libraries [1] to non-emscripten targets, such as
wasm32-wasi. Previously, only static linking was supported, and the-sharedand-fPICflags were simply ignored. Now both flags are honored.Since WASI runtimes do not necessarily include JavaScript support, we cannot rely on the JS-based Emscripten linker to link shared libraries. Instead, we link them using the Component Model proposal [2].
We have prototyped shared library support in
wasi-sdk[3] and put together a demo [4] which uses a patched version ofwit-component[5] to link libraries using the Component Model. We plan to submit the required changes upstream to the respective repositories in the next week or two.[1] https://github.com/WebAssembly/tool-conventions/blob/main/DynamicLinking.md
[2] https://github.com/WebAssembly/component-model/blob/main/design/mvp/examples/SharedEverythingDynamicLinking.md
[3] https://github.com/dicej/wasi-sdk/tree/dynamic-linking
[4] https://github.com/dicej/component-linking-demo
[5] https://github.com/bytecodealliance/wasm-tools/tree/main/crates/wit-component
Reviewed By: sbc100
Differential Revision: https://reviews.llvm.org/D153293