Remove `no_integrated_as` mode.
Specifically, remove both `-Z no_integrated_as` and
`TargetOptions::no_integrated_as`. The latter was only used for the
`msp430_none_elf` platform, for which it's no longer required.
r? @alexcrichton
use "gcc" instead of "cc" on *-sun-solaris systems when linking
On illumos and Solaris systems, Rust will use GCC as the link editor.
Rust does this by invoking "cc", which on many (Linux and perhaps BSD)
systems is generally either GCC or a GCC-compatible front-end. On
historical Solaris systems, "cc" was often the Sun Studio compiler.
This history casts a long shadow, and as such, even most modern
illumos-based operating systems tend to install GCC as "gcc", without
also making it available as "cc".
We should invoke GCC as "gcc" on such systems to ensure we get the right
compiler driver.
Specifically, remove both `-Z no_integrated_as` and
`TargetOptions::no_integrated_as`. The latter was only used for the
`msp430_none_elf` platform, for which it's no longer required.
Currently, there are three fields in `ModuleConfig` that dictate
how object files are emitted: `emit_obj`, `obj_is_bitcode`, and
`embed_bitcode`.
Some of the combinations of these fields are nonsensical, in particular
having both `obj_is_bitcode` and `embed_bitcode` true at the same time.
Also, currently:
- we needlessly emit and then delete a bytecode file if `obj_is_bitcode`
is true but `emit_obj` is false;
- we needlessly embed bitcode in the LLVM module if `embed_bitcode` is
true and `emit_obj` is false.
This commit combines the three fields into one, with a new type
`EmitObj` (and the auxiliary `BitcodeSection`) which can encode five
different possibilities.
In the old code, `set_flags` would set `obj_is_bitcode` and
`embed_bitcode` on all three of the configs (`modules`, `allocator`,
`metadata`) if the relevant other conditions were met, even if no object
code needed to be emitted for one or more of them. Whereas
`start_async_codegen` would set `emit_obj`, but only for those configs
that need it.
In the new code, `start_async_codegen` does all the work of setting
`emit_obj`, and it only does that for the configs that need it.
`set_flags` no longer sets anything related to object file emission.
Refactor `codegen`
`codegen` in `src/librustc_codegen_llvm/back/write.rs` is long and has complex control flow. These commits refactor it and make it easier to understand.
Refactorings to get rid of rustc_codegen_utils
r? @eddyb
cc #45276
After this, the only modules left in `rustc_codegen_utils` are
- `link`: a bunch of linking-related functions (many dealing with file names). These are mostly consumed by save analysis, rustc_driver, rustc_interface, and of course codegen. I assume they live here because we don't want a dependency of save analysis on codegen... Perhaps they can be moved to librustc?
- ~`symbol_names` and `symbol_names_test`: honestly it seems odd that `symbol_names_test` is not a submodule of `symbol_names`. It seems like these could honestly live in their own crate or move to librustc. Already name mangling is exported as the `symbol_name` query.~ (move it to its own crate)
I don't mind doing either of the above as part of this PR or a followup if you want.
Remove some imports to the rustc crate
- When we have `NestedVisitorMap::None`, we use `type Map = dyn intravisit::Map<'v>;` instead of the actual map. This doesn't actually result in dynamic dispatch (in the future we may want to use an associated type default to simplify the code).
- Use `rustc_session::` imports instead of `rustc::{session, lint}`.
r? @Zoxc
On illumos and Solaris systems, Rust will use GCC as the link editor.
Rust does this by invoking "cc", which on many (Linux and perhaps BSD)
systems is generally either GCC or a GCC-compatible front-end. On
historical Solaris systems, "cc" was often the Sun Studio compiler.
This history casts a long shadow, and as such, even most modern
illumos-based operating systems tend to install GCC as "gcc", without
also making it available as "cc".
We should invoke GCC as "gcc" on such systems to ensure we get the right
compiler driver.
Optimize catch_unwind to match C++ try/catch
This refactors the implementation of catching unwinds to allow LLVM to inline the "try" closure directly into the happy path, avoiding indirection. This means that the catch_unwind implementation is (after this PR) zero-cost unless a panic is thrown.
https://rust.godbolt.org/z/cZcUSB is an example of the current codegen in a simple case. Notably, the codegen is *exactly the same* if `-Cpanic=abort` is passed, which is clearly not great.
This PR, on the other hand, generates the following assembly:
```asm
# -Cpanic=unwind:
push rbx
mov ebx,0x2a
call QWORD PTR [rip+0x1c53c] # <happy>
mov eax,ebx
pop rbx
ret
mov rdi,rax
call QWORD PTR [rip+0x1c537] # cleanup function call
call QWORD PTR [rip+0x1c539] # <unfortunate>
mov ebx,0xd
mov eax,ebx
pop rbx
ret
# -Cpanic=abort:
push rax
call QWORD PTR [rip+0x20a1] # <happy>
mov eax,0x2a
pop rcx
ret
```
Fixes#64224, and resolves#64222.
cleanup more iterator usages (and other things)
* Improve weird formatting by moving comment inside else-code block.
* Use .any(x) instead of .find(x).is_some() on iterators.
* Use .nth(x) instead of .skip(x).next() on iterators.
* Simplify conditions like x + 1 <= y to x < y
* Use let instead of match to get value of enum with single variant.
Previously, the code responsible for handling the cycles between crates
introduces through weak lang items, would keep a set of missing language
items:
* extending it with items missing from the current crate,
* removing items provided by the current crate,
* grouping the crates when the set changed from non-empty back to empty.
This could produce incorrect results, if a lang item was missing from a
crate that comes after the crate that provides it (in the loop iteration
order). In that case the grouping would not take place.
The changes here address this specific failure scenario by keeping track
of two separate sets of crates. Those that are required to link successfully,
and those that are available for linking.
Verified using test case from 69368.
Support new LLVM pass manager
Add support for the new LLVM pass manager behind a `-Z new-llvm-pass-manager=on` option. Both the pre-link optimization and LTO pipelines use the new pass manager. There's some bits that are not supported yet:
* `-C passes`. NewPM requires an entirely different way of specifying custom pass pipelines. We should probably expose that functionality, but it doesn't directly map to what `-C passes` does.
* NewPM has no support for custom inline parameters right now. We'd have to add upstream support for that first.
* NewPM does not support PGO at O0 in LLVM 9 (which is why those tests fail with NewPM enabled). This is supported in LLVM 10.
* NewPM does not support MergeFunctions in LLVM 9. I've landed this upstream just before the cut, so we'll be able to re-enable that with LLVM 10.
Closes#64289.
r? @ghost
windows-gnu: prefer system crt libraries if they are available
The origin of the issue is the fact Rust ships mingw-w64 libraries but no headers and prefers own libraries over the system ones.
This leads to situation when headers aren't compatible with libraries (mingw-w64 doesn't provide any forward compatibility and AFAIK backwards compatibility is guaranteed only within major release series).
It's easier to understand how this PR works when looking at the linker invocation before and with this PR: https://www.diffchecker.com/GEuYFmzo
It adds system libraries path before Rust libraries so the linker will prefer them.
It has potential issue when system has files with the same names as Rust but that could be avoided by moving Rust shipped mingw-w64 libraries from `lib/rustlib/x86_64-pc-windows-gnu/lib` to say `lib/rustlib/x86_64-pc-windows-gnu/lib/mingw`. Then adding linker paths in this order: Rust libraries, system libraries, Rust shipped mingw-w64 libraries.
Fixes#47048Fixes#49078Fixes#53454Fixes#60912
Slimmer syntax
High-level summary of changes:
- The `syntax::node_count` pass is moved into `rustc_ast_passes`. This works towards improving #65031 by making compiling `syntax` go faster.
- The `syntax::{GLOBALS, with_globals, ..}` business is consolidated into `syntax::attr` for cleaner code and future possible improvements.
- The pretty printer loses its dependency on `ParseSess`, opting to use `SourceMap` & friends directly instead.
- Some drive by cleanup of `syntax::attr::HasAttr` happens.
- Builtin attribute logic (`syntax::attr::builtin`) + `syntax::attr::allow_internal_unstable` is moved into a new `rustc_attr` crate. More logic from `syntax::attr` should be moved into that crate over time. This also means that `syntax` loses all mentions of `ParseSess`, which enables the next point.
- The pretty printer `syntax::print` is moved into a new crate `rustc_ast_pretty`.
- `rustc_session::node_id` is moved back as `syntax::node_id`. As a result, `syntax` gets to drop dependencies on `rustc_session` (and implicitly `rustc_target`), `rustc_error_codes`, and `rustc_errors`. Moreover `rustc_hir` gets to drop its dependency on `rustc_session` as well. At this point, these crates are mostly "pure data crates", which is approaching a desirable end state.
- We should consider renaming `syntax` to `rustc_ast` now.
This patch enables rustc to emit the required LLVM module flags to enable Control Flow Guard metadata (cfguard=1) or metadata and checks (cfguard=2). The LLVM module flags are ignored on unsupported targets and operating systems.
Also share drop-glue when compiling with -Zshare-generics (i.e. at opt-level=0)
This PR adds drop-glue to the set of monomorphizations that can be shared across crates via `-Zshare-generics`.
This version of the PR might have detrimental effects on performance as it makes lots of stuff dependent on a single query results (`upstream_monomorphizations_for(def_id_of_drop_in_place)`). That should be fixable but let's do a perf run first.
Potentially fixes issue https://github.com/rust-lang/rust/issues/64140. (cc @alexcrichton)
The changes here are related to @matthewjasper's https://github.com/rust-lang/rust/pull/67332 but should be mostly orthogonal.
r? @ghost
Enable ASan on Fuchsia
This change adds the x86_64-fuchsia and aarch64-fuchsia LLVM targets to
those allowed to invoke -Zsanitizer. Currently, the only overlap between
compiler_rt sanitizers supported by both rustc and Fuchsia is ASan.
Export weak symbols used by MemorySanitizer
Export weak symbols defined by MemorySanitizer instrumentation, which are used
to implement `-Zsanitizer-memory-track-origins` and `-Zsanitizer-recover=memory`.
Previously, when using fat LTO, they would internalized and eliminated.
Fixes#68367.
This change adds the x86_64-fuchsia and aarch64-fuchsia LLVM targets to
those allowed to invoke -Zsanitizer. Currently, the only overlap between
compiler_rt sanitizers supported by both rustc and Fuchsia is ASan.
