Initialize default providers only once
This avoids copying a new `Providers` struct for each downstream crate
that wants to use it.
Follow-up to https://github.com/rust-lang/rust/pull/74283 without the perf hit.
r? @eddyb
Note that the output of `unpretty-debug.stdout` has changed. In that
test the hash values are normalized from a symbol numbers to small
numbers like "0#0" and "0#1". The increase in the number of static
symbols must have caused the original numbers to contain more digits,
resulting in different pretty-printing prior to normalization.
Eliminate confusing "globals" terminology.
There are some structures that are called "globals", but are they global
to a compilation session, and not truly global. I have always found this
highly confusing, so this commit renames them as "session globals" and
adds a comment explaining things.
Also, the commit fixes an unnecessary nesting of `set()` calls
`src/librustc_errors/json/tests.rs`
r? @Aaron1011
There are some structures that are called "globals", but are they global
to a compilation session, and not truly global. I have always found this
highly confusing, so this commit renames them as "session globals" and
adds a comment explaining things.
Also, the commit fixes an unnecessary nesting of `set()` calls
`src/librustc_errors/json/tests.rs`
Move remaining `NodeId` APIs from `Definitions` to `Resolver`
Implements https://github.com/rust-lang/rust/pull/73291#issuecomment-643515557
TL;DR: it moves all fields that are only needed during name resolution passes into the `Resolver` and keep the rest in `Definitions`. This effectively enforces that all references to `NodeId`s are gone once HIR lowering is completed.
After this, the only remaining work for #50928 should be to adjust the dev guide.
r? @petrochenkov
Fix -Z unpretty=everybody_loops
It turns out that this has not been working for who knows how long.
Previously:
```
pub fn h() { 1 + 2; }
```
After this change:
```
pub fn h() { loop { } }
```
This only affected the pass when run with the command line
pretty-printing option, so rustdoc was still replacing bodies with
`loop {}`.
Diagnose use of incompatible sanitizers
Emit an error when incompatible sanitizer are configured through command
line options. Previously the last one configured prevailed and others
were silently ignored.
Additionally use a set to represent configured sanitizers, making it
possible to enable multiple sanitizers at once. At least in principle,
since currently all of them are considered to be incompatible with
others.
It turns out that this has not been working for who knows how long.
Previously:
```
pub fn h() { 1 + 2; }
```
After this change:
```
pub fn h() { loop {} }
```
This only affected the pass when run with the command line
pretty-printing option, so rustdoc was still replacing bodies with
`loop {}`.
This initial version only injects counters at the top of each function.
Rust Coverage will require injecting additional counters at each
conditional code branch.
Emit an error when incompatible sanitizer are configured through command
line options. Previously the last one configured prevailed and others
were silently ignored.
Additionally use a set to represent configured sanitizers, making it
possible to enable multiple sanitizers at once. At least in principle,
since currently all of them are considered to be incompatible with
others.
expand: More precise locations for expansion-time lints
First commit: a macro expansion doesn't have a `NodeId` associated with it, but it has a parent `DefId` which we can use for linting.
The observable effect is that lints associated with macro expansions can now be `allow`ed at finer-grained level than whole crate.
Second commit: each macro definition has a `NodeId` which we can use for linting, unless that macro definition was decoded from other crate.
This is only really useful in debug messages, so I've switched to
calling `span_to_string` in any place that causes a `Span` to end up in
user-visible output.
Add `-Z span-debug` to allow for easier debugging of proc macros
Currently, the `Debug` impl for `proc_macro::Span` just prints out
the byte range. This can make debugging proc macros (either as a crate
author or as a compiler developer) very frustrating, since neither the
actual filename nor the `SyntaxContext` is displayed.
This commit adds a perma-unstable flag `-Z span-debug`. When enabled,
the `Debug` impl for `proc_macro::Span` simply forwards directly to
`rustc_span::Span`. Once #72618 is merged, this will start displaying
actual line numbers.
While `Debug` impls are not subject to Rust's normal stability
guarnatees, we probably shouldn't expose any additional information on
stable until `#![feature(proc_macro_span)]` is stabilized. Otherwise,
we would be providing a 'backdoor' way to access information that's
supposed be behind unstable APIs.
Currently, the `Debug` impl for `proc_macro::Span` just prints out
the byte range. This can make debugging proc macros (either as a crate
author or as a compiler developer) very frustrating, since neither the
actual filename nor the `SyntaxContext` is displayed.
This commit adds a perma-unstable flag `-Z span-debug`. When enabled,
the `Debug` impl for `proc_macro::Span` simply forwards directly to
`rustc_span::Span`. Once #72618 is merged, this will start displaying
actual line numbers.
While `Debug` impls are not subject to Rust's normal stability
guarnatees, we probably shouldn't expose any additional information on
stable until `#![feature(proc_macro_span)]` is stabilized. Otherwise,
we would be providing a 'backdoor' way to access information that's
supposed be behind unstable APIs.
Make `SourceMap` available for early debug-printing of `Span`s
Normally, we debug-print `Spans` using the `SourceMap` retrieved from
the global `TyCtxt`. However, we fall back to printing out the `Span`'s
raw fields (instead of a file and line number) when we try to print a
`Span` before a `TyCtxt` is available. This makes debugging early phases
of the compile, such as parsing, much more difficult.
This commit stores a `SourceMap` in `rustc_span::GlOBALS` as a fallback.
When a `TyCtxt` is not available, we try to retrieve one from `GLOBALS`
- only if this is not available do we fall back to the raw field output.
I'm not sure how to write a test for this - however, this can be
verified locally by setting `RUSTC_LOG="rustc_parse=debug"`, and
verifying that the output contains filenames and line numbers.
Normally, we debug-print `Spans` using the `SourceMap` retrieved from
the global `TyCtxt`. However, we fall back to printing out the `Span`'s
raw fields (instead of a file and line number) when we try to print a
`Span` before a `TyCtxt` is available. This makes debugging early phases
of the compile, such as parsing, much more difficult.
This commit stores a `SourceMap` in `rustc_span::GlOBALS` as a fallback.
When a `TyCtxt` is not available, we try to retrieve one from `GLOBALS`
- only if this is not available do we fall back to the raw field output.
I'm not sure how to write a test for this - however, this can be
verified locally by setting `RUSTC_LOG="rustc_parse=debug"`, and
verifying that the output contains filenames and line numbers.
Track devirtualized filenames
Split payload of FileName::Real to track both real and virtualized paths.
(Such splits arise from metadata refs into libstd; the virtualized paths look like `/rustc/1.45.0/src/libstd/io/cursor.rs` rather than `/Users/felixklock/Dev/Mozilla/rust.git/src/libstd/io/cursor.rs`)
This way, we can emit the virtual name into things like the like the StableSourceFileId (as was done back before PR #70642) that ends up in incremental build artifacts, while still using the devirtualized file path when we want to access the file.
Fix#70924
Add -Z profile-emit=<path> for Gcov gcda output.
Adds a -Z flag to control the file path that the Gcov gcda output is
written to during runtime. This flag expects a path and filename, e.g.
-Z profile-emit=gcov/out/lib.gcda.
This works similar to GCC/Clang's -fprofile-dir flag which allows
control over the output path for gcda coverage files.
Such splits arise from metadata refs into libstd.
This way, we can (in a follow on commit) continue to emit the virtual name into
things like the like the StableSourceFileId that ends up in incremetnal build
artifacts, while still using the devirtualized file path when we want to access
the file.
Note that this commit is intended to be a refactoring; the actual fix to the bug
in question is in a follow-on commit.
Adds a -Z flag to control the file path that the Gcov gcda output is
written to during runtime. This flag expects a path and filename, e.g.
-Z profile-emit=gcov/out/lib.gcda.
This works similar to GCC/Clang's -fprofile-dir flag which allows
control over the output path for gcda coverage files.
