Since LLVM reversed the order of the debug info graphs, we need to have
a compile unit that exists *before* any functions (`DISubprogram`s) are
created. This allows the LLVM debug info builder to automatically link
the functions to the compile unit.
In the older version, a `.o` and ` .bc` file were separate
work-products. This newer version keeps, for each codegen-unit, a set
of files of different kinds. We assume that if any kinds are available
then all the kinds we need are available, since the precise set of
switches will depend on attributes and command-line switches.
Should probably test this: the effect of changing attributes in
particular might not be successfully tracked?
This checks the `previous_work_products` data from the dep-graph and
tries to simply copy a `.o` file if possible. We also add new
work-products into the dep-graph, and create edges to/from the dep-node
for a work-product.
Centralize and clean type error reporting
Refactors the code that handles type errors to be cleaner and fixes various edge cases.
This made the already-bad "type mismatch resolving" error message somewhat uglier. I want to fix that in another commit before this PR is merged.
Fixes#31173
r? @jonathandturner, cc @nikomatsakis
Implement rust_eh_personality in Rust, remove rust_eh_personality_catch.
Well, not quite: ARM EHABI platforms still use the old scheme -- for now.
r? @alexcrichton
Refactor constant evaluation to use a single error reporting function
that reports a type-error-like message.
Also, unify all error codes with the "constant evaluation error" message
to just E0080, and similarly for a few other duplicate codes. The old
situation was a total mess, and now that we have *something* we can
further iterate on the UX.
The initial linkage-assignment (especially for closures) is a conservative one that makes some symbols more visible than they need to be. While this is not a correctness problem, it does force the LLVM inliner to be more conservative too, which results in poor performance. Once translation is based solely on MIR, it will be easier to also make the initial linkage assignment a better fitting one. Until then `internalize_symbols()` does a good job of preventing most performance regressions.
Simplify librustc_errors
This is part 2 of the error crate refactor, starting with #34403.
In this refactor, I focused on slimming down the error crate to fewer moving parts. As such, I've removed quite a few parts and replaced the with simpler, straight-line code. Specifically, this PR:
* Removes BasicEmitter
* Remove emit from emitter, leaving emit_struct
* Renames emit_struct to emit
* Removes CoreEmitter and focuses on a single Emitter
* Implements the latest changes to error format RFC (#1644)
* Removes (now-unused) code in emitter.rs and snippet.rs
* Moves more tests to the UI tester, removing some duplicate tests in the process
There is probably more that could be done with some additional refactoring, but this felt like it was getting to a good state.
r? @alexcrichton cc: @Manishearth (as there may be breaking changes in stuff I removed/changed)
Some more pattern cleanup and bugfixing
The next part of https://github.com/rust-lang/rust/pull/34095
The most significant fixed mistake is definitions for partially resolved associated types not being updated after full resolution.
```
fn f<T: Fn()>(arg: T::Output) { .... } // <- the definition of T::Output was not updated in def_map
```
For this reason unstable associated types of stable traits, like `FnOnce::Output`, could be used in stable code when written in unqualified form. Now they are properly checked, this is a **[breaking-change]** (pretty minor one, but a crater run would be nice). The fix is not to use unstable library features in stable code, alternatively `FnOnce::Output` can be stabilized.
Besides that, paths in struct patterns and expressions `S::A { .. }` are now fully resolved as associated types. Such types cannot be identified as structs at the moment, i.e. the change doesn't make previously invalid code valid, but it improves error diagnostics.
Other changes: `Def::Err` is supported better (less chances for ICEs for erroneous code), some incorrect error messages are corrected, some duplicated error messages are not reported, ADT definitions are now available through constructor IDs, everything else is cleanup and code audit.
Fixes https://github.com/rust-lang/rust/issues/34209
Closes https://github.com/rust-lang/rust/issues/22933 (adds tests)
r? @eddyb
The data tracked here was meant to compare the output of the
translation item collector to the set of translation items found
by the on-demand translator.
Functions and method are declared ahead-of-time, including generic ones.
Closures are not considered trans items anymore, instead they are
translated on demands.
