Implement Unsized Rvalues
This PR is the first step to implement RFC1909: unsized rvalues (#48055).
## Implemented
- `Sized` is removed for arguments and local bindings. (under `#![feature(unsized_locals)]`)
- Unsized locations are allowed in MIR
- Unsized places and operands are correctly translated at codegen
## Not implemented in this PR
- Additional `Sized` checks:
- tuple struct constructor (accidentally compiles now)
- closure arguments at closure generation (accidentally compiles now)
- upvars (ICEs now)
- Generating vtable for `fn method(self)` (ICEs now)
- VLAs: `[e; n]` where `n` isn't const
- Reduce unnecessary allocations
## Current status
- [x] Fix `__rust_probestack` (rust-lang-nursery/compiler-builtins#244)
- [x] Get the fix merged
- [x] `#![feature(unsized_locals)]`
- [x] Give it a tracking issue number
- [x] Lift sized checks in typeck and MIR-borrowck
- [ ] <del>Forbid `A(unsized-expr)`</del> will be another PR
- [x] Minimum working codegen
- [x] Add more examples and fill in unimplemented codegen paths
- [ ] <del>Loosen object-safety rules (will be another PR)</del>
- [ ] <del>Implement `Box<FnOnce>` (will be another PR)</del>
- [ ] <del>Reduce temporaries (will be another PR)</del>
Set more llvm function attributes for __rust_try
This shim is generated elsewhere in the compiler so this commit adds support to
ensure it goes through similar paths as the rest of the compiler to set llvm
function attributes like target features.
cc #53372
This shim is generated elsewhere in the compiler so this commit adds support to
ensure it goes through similar paths as the rest of the compiler to set llvm
function attributes like target features.
cc #53372
Originally added in #52887 this commit disables passing `--strip-debug` to LLD
when optimized. This bring back the original bug of emitting broken debuginfo
but currently it *also* strips the `name` section which makes it very difficult
to inspect the final binary. A real fix is happening at
https://reviews.llvm.org/D50729 and we can reevaluate once we've updated LLD to
have that commit.
rustc_codegen_llvm: Restore the closure env alloca hack for LLVM 5.
This hack was removed in #50949, but without it I found that building
`std` with full debuginfo would print many LLVM `DW_OP_LLVM_fragment`
errors, then die `LLVM ERROR: Failed to strip malformed debug info`.
It doesn't seem to be a problem for LLVM 6, so we can re-enable the hack
just for older LLVM.
This reverts commit da579ef75e.
Fixes#53204.
r? @eddyb
As explained by eddyb in #53221, "An &ArchiveChild doesn't point into the archive itself, it points to an owned object that itself points to the archive, and LLVMRustArchiveMemberNew copies the ArchiveChild (whereas the current signature suggests it keeps the &ArchiveChild)."
Cleanup to librustc::session and related code
No functional changes, just some cleanup.
This also creates the `rustc_fs_util` crate, but I can remove that change if desired. It felt a little odd to force crates to depend on librustc for some fs utilities; and also seemed good to generally keep the size of librustc lower (for compile times); fs_util will compile in parallel with essentially the first crate since it has no dependencies beyond std.
This hack was removed in #50949, but without it I found that building
`std` with full debuginfo would print many LLVM `DW_OP_LLVM_fragment`
errors, then die `LLVM ERROR: Failed to strip malformed debug info`.
It doesn't seem to be a problem for LLVM 6, so we can re-enable the hack
just for older LLVM.
This reverts commit da579ef75e.
Fixes#53204.
r? @eddyb
Rollup of 15 pull requests
Successful merges:
- #52773 (Avoid unnecessary pattern matching against Option and Result)
- #53082 (Fix doc link (again))
- #53094 (Automatically expand section if url id point to one of its component)
- #53106 (atomic ordering docs)
- #53110 (Account for --remap-path-prefix in save-analysis)
- #53116 (NetBSD: fix signedess of char)
- #53179 (Whitelist wasm32 simd128 target feature)
- #53183 (Suggest comma when missing in macro call)
- #53207 (Add individual docs for rotate_{left, right})
- #53211 ([nll] enable feature(nll) on various crates for bootstrap)
- #53214 ([nll] enable feature(nll) on various crates for bootstrap: part 2)
- #53215 (Slightly refactor syntax_ext/format)
- #53217 (inline some short functions)
- #53219 ([nll] enable feature(nll) on various crates for bootstrap: part 3)
- #53222 (A few cleanups for rustc_target)
Apply some fixes to cross-language LTO (especially when targeting MSVC)
This PR contains a few fixes that were needed in order to get Firefox compiling with Rust/C++ cross-language ThinLTO on Windows. The commits are self-contained and should be self-explanatory.
r? @alexcrichton
This commit tweaks the linker-level visibility of some lang items that rustc
uses and defines. Notably this means that `#[panic_implementation]` and
`#[alloc_error_handler]` functions are never marked as `internal`. It's up to
the linker to eliminate these, not rustc.
Additionally `#[global_allocator]` generated symbols are no longer forced to
`Default` visibility (fully exported), but rather they're relaxed to `Hidden`
visibility). This symbols are *not* needed across DLL boundaries, only as a
local implementation detail of the compiler-injected allocator symbols, so
`Hidden` should suffice.
Closes#51342Closes#52795
with ThinLTO and cross-lang-lto.
Normally, when compiling with whole-crate-graph ThinLTO, we expect
rustc's LTO step to "uplift" upstream object files/LLVM modules to
the current set of compilation artifacts. Therefore the staticlib
creation code skips this uplifting. However, when compiling with
"cross-language LTO" (i.e. defer LTO to the actual linker), the LTO
step in rustc is not performed, so we have to take care of copying
upstream object files during archive creation (like we already do
when compiling without any LTO).
Building librustc_codegen_llvm in a separate directory
This allows clearing it out and building it separately from the
compiler. Since it's essentially a different and separate crate this
makes sense to do, each cargo invocation should generally happen in its
own directory.
r? @alexcrichton
This allows clearing it out and building it separately from the
compiler. Since it's essentially a different and separate crate this
makes sense to do, each cargo invocation should generally happen in its
own directory.
Reintroduce `Undef` and properly check constant value sizes
r? @RalfJung
cc @eddyb
basically all kinds of silent failures that never occurred are assertions now
Disable debug sections when optimization flags is set for LLD.
Currently LLD does not error when optimization is set and debugging information sections are present. (See discussion at https://reviews.llvm.org/D47901)
Using `--strip-debug` along with the `-O` option.
Previously linker diagnostic were being hidden when two modules were linked
together but failed to link. This commit fixes the situation by ensuring that we
have a diagnostic handler installed and also adds support for handling linker
diagnostics.
Don't format!() string literals
Prefer `to_string()` to `format!()` take 2, this time targetting string literals. In some cases (`&format!("...")` -> `"..."`) also removes allocations. Occurences of `format!("")` are changed to `String::new()`.
Replace push loops with extend() where possible
Or set the vector capacity where I couldn't do it.
According to my [simple benchmark](https://gist.github.com/ljedrz/568e97621b749849684c1da71c27dceb) `extend`ing a vector can be over **10 times** faster than `push`ing to it in a loop:
10 elements (6.1 times faster):
```
test bench_extension ... bench: 75 ns/iter (+/- 23)
test bench_push_loop ... bench: 458 ns/iter (+/- 142)
```
100 elements (11.12 times faster):
```
test bench_extension ... bench: 87 ns/iter (+/- 26)
test bench_push_loop ... bench: 968 ns/iter (+/- 3,528)
```
1000 elements (11.04 times faster):
```
test bench_extension ... bench: 311 ns/iter (+/- 9)
test bench_push_loop ... bench: 3,436 ns/iter (+/- 233)
```
Seems like a good idea to use `extend` as much as possible.
Improve a few vectors - calculate capacity or build from iterators
Collecting from iterators improves readability and tailoring vector capacities should be beneficial in terms of performance.
Rollup of bare_trait_objects PRs
All deny attributes were moved into bootstrap so they can be disabled with a line of config.
Warnings for external tools are allowed and it's up to the tool's maintainer to keep it warnings free.
r? @Mark-Simulacrum
cc @ljedrz @kennytm
Rollup of 16 pull requests
Successful merges:
- #52558 (Add tests for ICEs which no longer repro)
- #52610 (Clarify what a task is)
- #52617 (Don't match on region kinds when reporting NLL errors)
- #52635 (Fix #[linkage] propagation though generic functions)
- #52647 (Suggest to take and ignore args while closure args count mismatching)
- #52649 (Point spans to inner elements of format strings)
- #52654 (Format linker args in a way that works for gcc and ld)
- #52667 (update the stdsimd submodule)
- #52674 (Impl Executor for Box<E: Executor>)
- #52690 (ARM: expose `rclass` and `dsp` target features)
- #52692 (Improve readability in a few sorts)
- #52695 (Hide some lints which are not quite right the way they are reported to the user)
- #52718 (State default capacity for BufReader/BufWriter)
- #52721 (std::ops::Try impl for std::task::Poll)
- #52723 (rustc: Register crates under their real names)
- #52734 (sparc ABI issue - structure returning from function is returned in 64bit registers (with tests))
Failed merges:
- #52678 ([NLL] Use better spans in some errors)
r? @ghost
introduce universes to NLL type check
This branch aims to fix#48071 and also advance chalk integration a bit at the same time. It re-implements the subtyping/type-equating check so that NLL doesn't "piggy back" on the subtyping code of the old type checker.
This new code uses the "universe-based" approach to handling higher-ranked lifetimes, which sidesteps some of the limitations of the current "leak-based" scheme. This avoids the ICE in #48071.
At the same time, I aim for this to potentially be a kind of optimization. This NLL code is (currently) not cached, but it also generates constraints without doing as much instantiation, substitution, and folding. Right now, though, it still piggy backs on the `relate_tys` trait, which is a bit unfortunate -- it means we are doing more hashing and things than we have to. I want to measure the see the perf. Refactoring that trait is something I'd prefer to leave for follow-up work.
r? @pnkfelix -- but I want to measure perf etc first
ARM: expose `rclass` and `dsp` target features
- `dsp`: the subtarget supports the DSP (saturating arith. and such)
instructions
- `rclass`: target is a Cortex-R
Both features are useful to support ARM MCUs on `coresimd`.
Note: Cortex-R52 is the first Armv8-R with `neon` support.
r? @alexcrichton
cc @japaric
Format linker args in a way that works for gcc and ld
Pass multiple linker arguments rather than concatenate with commas (fixes#52634).
`-l library` -> `-llibrary` to work with apple's ld.
To build with apple's ld I'm currently also passing `-C link-args="-arch x86_64 -macosx_version_min 10.13.0"`. I'll try and understand the latter flag better before PRing that.
This PR currently works for me. Hopefully CI will pick up any grievous ramifications in other toolchains?
Thanks to @alexcrichton for the pointer to the relevant code!
Fix #[linkage] propagation though generic functions
Fixes#18804
In the non-local branch of `get_static` (where the fix was implemented) `span_fatal` had to be replaced with `bug!` as we have no span in that case.
- `dsp`: the subtarget supports the DSP (saturating arith. and such)
instructions
- `rclass`: target is a Cortex-R
Both features are useful to support ARM MCUs on `coresimd`.
Note: Cortex-R52 is the first Armv8-R with `neon` support
Add unaligned volatile intrinsics
Surprisingly enough, it turns out that unaligned volatile loads are actually useful for certain (very niche) types of lock-free code. I included unaligned volatile stores for completeness, but I currently do not know of any use cases for them.
These are only exposed as intrinsics for now. If they turn out to be useful in practice, we can work towards stabilizing them.
r? @alexcrichton
Calculate Vec capacities in librustc
Calculate the required capacity of a few vectors in rustc based on the number of elements they are populated with.
rustc: Remove a workaround in ThinLTO fixed upstream
This commit removes a hack in our ThinLTO passes which removes available
externally functions manually. The [upstream bug][1] has long since been fixed,
so we should be able to rely on LLVM natively for this now!
[1]: https://bugs.llvm.org/show_bug.cgi?id=35736
This commit removes a hack in our ThinLTO passes which removes available
externally functions manually. The [upstream bug][1] has long since been fixed,
so we should be able to rely on LLVM natively for this now!
[1]: https://bugs.llvm.org/show_bug.cgi?id=35736
This commit stabilizes the `#[wasm_import_module]` attribute as
`#[link(wasm_import_module = "...")]`. Tracked by #52090 this new directive in
the `#[link]` attribute is used to configured the module name that the imports
are listed with. The WebAssembly specification indicates two utf-8 names are
associated with all imported items, one for the module the item comes from and
one for the item itself. The item itself is configurable in Rust via its
identifier or `#[link_name = "..."]`, but the module name was previously not
configurable and defaulted to `"env"`. This commit ensures that this is also
configurable.
Closes#52090
This commit transitions definitions of custom sections on the wasm target from
the unstable `#[wasm_custom_section]` attribute to the
already-stable-for-other-targets `#[link_section]` attribute. Mostly the same
restrictions apply as before, except that this now applies only to statics.
Closes#51088
