Add a min_atomic_width target option, like max_atomic_width
Rationale: some ISAs, e.g. OR1K, do not have atomic instructions
for byte and halfword access, and at the same time do not have
a fixed endianness, which makes it unreasonable to implement these
through word-sized atomic accesses.
Remove special case for Box<ZST> in trans
Remove extra lang item, `exchange_free`; use `box_free` instead.
Trans used to insert code equivalent to `box_free` in a wrapper around
`exchange_free`, and that code is now removed from trans.
Fixes#37710.
Use `DefId`s instead of `NodeId`s for `pub(restricted)` visibilities
This is groundwork for hygiene 2.0, specifically privacy checking hygienic intercrate name resolutions.
r? @nrc
Rationale: some ISAs, e.g. OR1K, do not have atomic instructions
for byte and halfword access, and at the same time do not have
a fixed endianness, which makes it unreasonable to implement these
through word-sized atomic accesses.
Disable field reordering
This was decided via IRC and needs a backport to beta. Basically, #37429 broke servo, and probably needs an announcement and opt-in flag. I didn't run all tests locally but think I've already reverted all the ones that need to be reverted.
r? @nikomatsakis
Redox Cross Compilation
I will admit - there are things here that I wish I did not have to do. This completes the ability to create a cross compiler from the rust repository for `x86_64-unknown-redox`. I will document this PR with inline comments explaining some things.
[View this gist to see how a cross compiler is built](https://gist.github.com/jackpot51/6680ad973986e84d69c79854249f2b7e)
Prior discussion of a smaller change is here: https://github.com/rust-lang/rust/pull/38366
Refactor global paths
This PR removes the field `global: bool` from `ast::Path` and `hir::Path`, instead representing a global path `::foo::bar` as `{{root}}::foo::bar`, where `{{root}}` is a virtual keyword `keywords::CrateRoot`.
Also, fixes#38016.
r? @nrc
Cleanup refactoring around DefPath handling
This PR makes two big changes:
* All DefPaths of a crate are now stored in metadata in their own table (as opposed to `DefKey`s as part of metadata `Entry`s.
* The compiler will no longer allocate a pseudo-local DefId for inlined HIR nodes (because those are gross). Inlined HIR nodes will have a NodeId but they don't have there own DefId anymore. Turns out they were not needed anymore either. Hopefully HIR inlining will be gone completely one day but if until then we start needing to be able to map inlined NodeIds to original DefIds, we can add an additional table to metadata that allows for reconstructing this.
Overall this makes for some nice simplifications and removal of special cases.
r? @eddyb
cc @rust-lang/compiler
Expand E0309 explanation with motivating example
I recently started reading @Gankro's "[Learning Rust With Entirely Too Many Linked Lists](http://cglab.ca/~abeinges/blah/too-many-lists/book/README.html)", and came across [a part](http://cglab.ca/~abeinges/blah/too-many-lists/book/second-iter.html) where he comes across `E0309`, and after showing `rustc --explain E0309` prompty says
> This is dumb. I think it's dumb. You have to do it.
Humor aside, I think this says something about the current explanation being somewhat lacking.
This patch introduces a motivating example saying why `T: 'a` is a necessary restriction. Hopefully, this will help new Rustaceans understand why leaving out the `'a` bound on `T` might lead to broken code.
Point out the known type when field doesn't satisfy bound
For file
```rust
use std::path::Path;
fn f(p: Path) { }
```
provide the following error
```nocode
error[E0277]: the trait bound `[u8]: std::marker::Sized` is not satisfied in `std::path::Path`
--> file.rs:3:6
|
3 | fn f(p: Path) { }
| ^ within `std::path::Path`, the trait `std::marker::Sized` is not implemented for `[u8]`
|
= note: `[u8]` does not have a constant size known at compile-time
= note: required because it appears within the type `std::path::Path`
= note: all local variables must have a statically known size
```
Fix#23286.
struct field reordering and optimization
This is work in progress. The goal is to divorce the order of fields in source code from the order of fields in the LLVM IR, then optimize structs (and tuples/enum variants)by always ordering fields from least to most aligned. It does not work yet. I intend to check compiler memory usage as a benchmark, and a crater run will probably be required.
I don't know enough of the compiler to complete this work unaided. If you see places that still need updating, please mention them. The only one I know of currently is debuginfo, which I'm putting off intentionally until a bit later.
r? @eddyb
Currently libraries installed by rustbuild on OSX have an incorrect
`LC_ID_DYLIB` directive located in the dynamic libraries that are
installed. The directive we expect looks like:
@rpath/libstd.dylib
Which means that if you want to find that dynamic library you should
look at the dylib's other `@rpath` directives. Typically our `@rpath`
directives look like `@loader_path/../lib` for the compiler as that's
where the installed libraries will be located. Currently, though,
rustbuild produces dylibs with the directive that looks like:
/Users/rustbuild/src/rust-buildbot/slave/nightly-dist-rustc-mac/build/build/x86_64-apple-darwin/stage1-std/x86_64-apple-darwin/release/deps/libstd-713ad88203512705.dylib
In other words, the build directory is encoded erroneously. The compiler
already [knows how] to change this directive, but it only passes that
argument when `-C rpath` is also passed. The rustbuild system, however,
explicitly [does not pass] this option explicitly and instead bakes its
own. This logic then also erroneously didn't pass `-Wl,-install_name`
like the compiler.
[knows how]: 4a008cccaa/src/librustc_trans/back/linker.rs (L210-L214)
[does not pass]: 4a008cccaa/src/bootstrap/bin/rustc.rs (L133-L158)
To fix this regression this patch introduces a new `-Z` flag, `-Z
osx-rpath-install-name` which basically just forces the compiler to take
the previous `-install_name` branch when creating a dynamic library.
Hopefully we can sort out a better rpath story in the future, but for
now this "hack" should suffice in getting our nightly builds back to the
same state as before.
Closes#38430
