Encode spans relative to the enclosing item
The aim of this PR is to avoid recomputing queries when code is moved without modification.
MCP at https://github.com/rust-lang/compiler-team/issues/443
This is achieved by :
1. storing the HIR owner LocalDefId information inside the span;
2. encoding and decoding spans relative to the enclosing item in the incremental on-disk cache;
3. marking a dependency to the `source_span(LocalDefId)` query when we translate a span from the short (`Span`) representation to its explicit (`SpanData`) representation.
Since all client code uses `Span`, step 3 ensures that all manipulations
of span byte positions actually create the dependency edge between
the caller and the `source_span(LocalDefId)`.
This query return the actual absolute span of the parent item.
As a consequence, any source code motion that changes the absolute byte position of a node will either:
- modify the distance to the parent's beginning, so change the relative span's hash;
- dirty `source_span`, and trigger the incremental recomputation of all code that
depends on the span's absolute byte position.
With this scheme, I believe the dependency tracking to be accurate.
For the moment, the spans are marked during lowering.
I'd rather do this during def-collection,
but the AST MutVisitor is not practical enough just yet.
The only difference is that we attach macro-expanded spans
to their expansion point instead of the macro itself.
Display an extra note for trailing semicolon lint with trailing macro
Currently, we parse macros at the end of a block
(e.g. `fn foo() { my_macro!() }`) as expressions, rather than
statements. This means that a macro invoked in this position
cannot expand to items or semicolon-terminated expressions.
In the future, we might want to start parsing these kinds of macros
as statements. This would make expansion more 'token-based'
(i.e. macro expansion behaves (almost) as if you just textually
replaced the macro invocation with its output). However,
this is a breaking change (see PR #78991), so it will require
further discussion.
Since the current behavior will not be changing any time soon,
we need to address the interaction with the
`SEMICOLON_IN_EXPRESSIONS_FROM_MACROS` lint. Since we are parsing
the result of macro expansion as an expression, we will emit a lint
if there's a trailing semicolon in the macro output. However, this
results in a somewhat confusing message for users, since it visually
looks like there should be no problem with having a semicolon
at the end of a block
(e.g. `fn foo() { my_macro!() }` => `fn foo() { produced_expr; }`)
To help reduce confusion, this commit adds a note explaining
that the macro is being interpreted as an expression. Additionally,
we suggest adding a semicolon after the macro *invocation* - this
will cause us to parse the macro call as a statement. We do *not*
use a structured suggestion for this, since the user may actually
want to remove the semicolon from the macro definition (allowing
the block to evaluate to the expression produced by the macro).
Currently, we parse macros at the end of a block
(e.g. `fn foo() { my_macro!() }`) as expressions, rather than
statements. This means that a macro invoked in this position
cannot expand to items or semicolon-terminated expressions.
In the future, we might want to start parsing these kinds of macros
as statements. This would make expansion more 'token-based'
(i.e. macro expansion behaves (almost) as if you just textually
replaced the macro invocation with its output). However,
this is a breaking change (see PR #78991), so it will require
further discussion.
Since the current behavior will not be changing any time soon,
we need to address the interaction with the
`SEMICOLON_IN_EXPRESSIONS_FROM_MACROS` lint. Since we are parsing
the result of macro expansion as an expression, we will emit a lint
if there's a trailing semicolon in the macro output. However, this
results in a somewhat confusing message for users, since it visually
looks like there should be no problem with having a semicolon
at the end of a block
(e.g. `fn foo() { my_macro!() }` => `fn foo() { produced_expr; }`)
To help reduce confusion, this commit adds a note explaining
that the macro is being interpreted as an expression. Additionally,
we suggest adding a semicolon after the macro *invocation* - this
will cause us to parse the macro call as a statement. We do *not*
use a structured suggestion for this, since the user may actually
want to remove the semicolon from the macro definition (allowing
the block to evaluate to the expression produced by the macro).
When we need to emit a lint at a macro invocation, we currently use the
`NodeId` of its parent definition (e.g. the enclosing function). This
means that any `#[allow]` / `#[deny]` attributes placed 'closer' to the
macro (e.g. on an enclosing block or statement) will have no effect.
This commit computes a better `lint_node_id` in `InvocationCollector`.
When we visit/flat_map an AST node, we assign it a `NodeId` (earlier
than we normally would), and store than `NodeId` in current
`ExpansionData`. When we collect a macro invocation, the current
`lint_node_id` gets cloned along with our `ExpansionData`, allowing it
to be used if we need to emit a lint later on.
This improves the handling of `#[allow]` / `#[deny]` for
`SEMICOLON_IN_EXPRESSIONS_FROM_MACROS` and some `asm!`-related lints.
The 'legacy derive helpers' lint retains its current behavior
(I've inlined the now-removed `lint_node_id` function), since
there isn't an `ExpansionData` readily available.
Fix `--remap-path-prefix` not correctly remapping `rust-src` component paths and unify handling of path mapping with virtualized paths
This PR fixes#73167 ("Binaries end up containing path to the rust-src component despite `--remap-path-prefix`") by preventing real local filesystem paths from reaching compilation output if the path is supposed to be remapped.
`RealFileName::Named` introduced in #72767 is now renamed as `LocalPath`, because this variant wraps a (most likely) valid local filesystem path.
`RealFileName::Devirtualized` is renamed as `Remapped` to be used for remapped path from a real path via `--remap-path-prefix` argument, as well as real path inferred from a virtualized (during compiler bootstrapping) `/rustc/...` path. The `local_path` field is now an `Option<PathBuf>`, as it will be set to `None` before serialisation, so it never reaches any build output. Attempting to serialise a non-`None` `local_path` will cause an assertion faliure.
When a path is remapped, a `RealFileName::Remapped` variant is created. The original path is preserved in `local_path` field and the remapped path is saved in `virtual_name` field. Previously, the `local_path` is directly modified which goes against its purpose of "suitable for reading from the file system on the local host".
`rustc_span::SourceFile`'s fields `unmapped_path` (introduced by #44940) and `name_was_remapped` (introduced by #41508 when `--remap-path-prefix` feature originally added) are removed, as these two pieces of information can be inferred from the `name` field: if it's anything other than a `FileName::Real(_)`, or if it is a `FileName::Real(RealFileName::LocalPath(_))`, then clearly `name_was_remapped` would've been false and `unmapped_path` would've been `None`. If it is a `FileName::Real(RealFileName::Remapped{local_path, virtual_name})`, then `name_was_remapped` would've been true and `unmapped_path` would've been `Some(local_path)`.
cc `@eddyb` who implemented `/rustc/...` path devirtualisation
This PR implements span quoting, allowing proc-macros to produce spans
pointing *into their own crate*. This is used by the unstable
`proc_macro::quote!` macro, allowing us to get error messages like this:
```
error[E0412]: cannot find type `MissingType` in this scope
--> $DIR/auxiliary/span-from-proc-macro.rs:37:20
|
LL | pub fn error_from_attribute(_args: TokenStream, _input: TokenStream) -> TokenStream {
| ----------------------------------------------------------------------------------- in this expansion of procedural macro `#[error_from_attribute]`
...
LL | field: MissingType
| ^^^^^^^^^^^ not found in this scope
|
::: $DIR/span-from-proc-macro.rs:8:1
|
LL | #[error_from_attribute]
| ----------------------- in this macro invocation
```
Here, `MissingType` occurs inside the implementation of the proc-macro
`#[error_from_attribute]`. Previosuly, this would always result in a
span pointing at `#[error_from_attribute]`
This will make many proc-macro-related error message much more useful -
when a proc-macro generates code containing an error, users will get an
error message pointing directly at that code (within the macro
definition), instead of always getting a span pointing at the macro
invocation site.
This is implemented as follows:
* When a proc-macro crate is being *compiled*, it causes the `quote!`
macro to get run. This saves all of the sapns in the input to `quote!`
into the metadata of *the proc-macro-crate* (which we are currently
compiling). The `quote!` macro then expands to a call to
`proc_macro::Span::recover_proc_macro_span(id)`, where `id` is an
opaque identifier for the span in the crate metadata.
* When the same proc-macro crate is *run* (e.g. it is loaded from disk
and invoked by some consumer crate), the call to
`proc_macro::Span::recover_proc_macro_span` causes us to load the span
from the proc-macro crate's metadata. The proc-macro then produces a
`TokenStream` containing a `Span` pointing into the proc-macro crate
itself.
The recursive nature of 'quote!' can be difficult to understand at
first. The file `src/test/ui/proc-macro/quote-debug.stdout` shows
the output of the `quote!` macro, which should make this eaier to
understand.
This PR also supports custom quoting spans in custom quote macros (e.g.
the `quote` crate). All span quoting goes through the
`proc_macro::quote_span` method, which can be called by a custom quote
macro to perform span quoting. An example of this usage is provided in
`src/test/ui/proc-macro/auxiliary/custom-quote.rs`
Custom quoting currently has a few limitations:
In order to quote a span, we need to generate a call to
`proc_macro::Span::recover_proc_macro_span`. However, proc-macros
support renaming the `proc_macro` crate, so we can't simply hardcode
this path. Previously, the `quote_span` method used the path
`crate::Span` - however, this only works when it is called by the
builtin `quote!` macro in the same crate. To support being called from
arbitrary crates, we need access to the name of the `proc_macro` crate
to generate a path. This PR adds an additional argument to `quote_span`
to specify the name of the `proc_macro` crate. Howver, this feels kind
of hacky, and we may want to change this before stabilizing anything
quote-related.
Additionally, using `quote_span` currently requires enabling the
`proc_macro_internals` feature. The builtin `quote!` macro
has an `#[allow_internal_unstable]` attribute, but this won't work for
custom quote implementations. This will likely require some additional
tricks to apply `allow_internal_unstable` to the span of
`proc_macro::Span::recover_proc_macro_span`.
This PR modifies the macro expansion infrastructure to handle attributes
in a fully token-based manner. As a result:
* Derives macros no longer lose spans when their input is modified
by eager cfg-expansion. This is accomplished by performing eager
cfg-expansion on the token stream that we pass to the derive
proc-macro
* Inner attributes now preserve spans in all cases, including when we
have multiple inner attributes in a row.
This is accomplished through the following changes:
* New structs `AttrAnnotatedTokenStream` and `AttrAnnotatedTokenTree` are introduced.
These are very similar to a normal `TokenTree`, but they also track
the position of attributes and attribute targets within the stream.
They are built when we collect tokens during parsing.
An `AttrAnnotatedTokenStream` is converted to a regular `TokenStream` when
we invoke a macro.
* Token capturing and `LazyTokenStream` are modified to work with
`AttrAnnotatedTokenStream`. A new `ReplaceRange` type is introduced, which
is created during the parsing of a nested AST node to make the 'outer'
AST node aware of the attributes and attribute target stored deeper in the token stream.
* When we need to perform eager cfg-expansion (either due to `#[derive]` or `#[cfg_eval]`),
we tokenize and reparse our target, capturing additional information about the locations of
`#[cfg]` and `#[cfg_attr]` attributes at any depth within the target.
This is a performance optimization, allowing us to perform less work
in the typical case where captured tokens never have eager cfg-expansion run.
Extend `proc_macro_back_compat` lint to `procedural-masquerade`
We now lint on *any* use of `procedural-masquerade` crate. While this
crate still exists, its main reverse dependency (`cssparser`) no longer
depends on it. Any crates still depending off should stop doing so, as
it only exists to support very old Rust versions.
If a crate actually needs to support old versions of rustc via
`procedural-masquerade`, then they'll just need to accept the warning
until we remove it entirely (at the same time as the back-compat hack).
The latest version of `procedural-masquerade` does work with the
latest rustc, but trying to check for the version seems like more
trouble than it's worth.
While working on this, I realized that the `proc-macro-hack` check was
never actually doing anything. The corresponding enum variant in
`proc-macro-hack` is named `Value` or `Nested` - it has never been
called `Input`. Due to a strange Crater issue, the Crater run that
tested adding this did *not* end up testing it - some of the crates that
would have failed did not actually have their tests checked, making it
seem as though the `proc-macro-hack` check was working.
The Crater issue is being discussed at
https://rust-lang.zulipchat.com/#narrow/stream/242791-t-infra/topic/Nearly.20identical.20Crater.20runs.20processed.20a.20crate.20differently/near/230406661
Despite the `proc-macro-hack` check not actually doing anything, we
haven't gotten any reports from users about their build being broken.
I went ahead and removed it entirely, since it's clear that no one is
being affected by the `proc-macro-hack` regression in practice.
StructField -> FieldDef ("field definition")
Field -> ExprField ("expression field", not "field expression")
FieldPat -> PatField ("pattern field", not "field pattern")
Also rename visiting and other methods working on them.
We now lint on *any* use of `procedural-masquerade` crate. While this
crate still exists, its main reverse dependency (`cssparser`) no longer
depends on it. Any crates still depending off should stop doing so, as
it only exists to support very old Rust versions.
If a crate actually needs to support old versions of rustc via
`procedural-masquerade`, then they'll just need to accept the warning
until we remove it entirely (at the same time as the back-compat hack).
The latest version of `procedural-masquerade` does not work with the
latest rustc, but trying to check for the version seems like more
trouble than it's worth.
While working on this, I realized that the `proc-macro-hack` check was
never actually doing anything. The corresponding enum variant in
`proc-macro-hack` is named `Value` or `Nested` - it has never been
called `Input`. Due to a strange Crater issue, the Crater run that
tested adding this did *not* end up testing it - some of the crates that
would have failed did not actually have their tests checked, making it
seem as though the `proc-macro-hack` check was working.
The Crater issue is being discussed at
https://rust-lang.zulipchat.com/#narrow/stream/242791-t-infra/topic/Nearly.20identical.20Crater.20runs.20processed.20a.20crate.20differently/near/230406661
Despite the `proc-macro-hack` check not actually doing anything, we
haven't gotten any reports from users about their build being broken.
I went ahead and removed it entirely, since it's clear that no one is
being affected by the `proc-macro-hack` regression in practice.
expand: Do not allocate `Lrc` for `allow_internal_unstable` list unless necessary
This allocation is done for any macro defined in the current crate, or used from a different crate.
EDIT: This also removes an `Lrc` increment from each *use* of such macro, which may be more significant.
Noticed when reviewing https://github.com/rust-lang/rust/pull/82367.
This probably doesn't matter, but let's do a perf run.
Implement built-in attribute macro `#[cfg_eval]` + some refactoring
This PR implements a built-in attribute macro `#[cfg_eval]` as it was suggested in https://github.com/rust-lang/rust/pull/79078 to avoid `#[derive()]` without arguments being abused as a way to configure input for other attributes.
The macro is used for eagerly expanding all `#[cfg]` and `#[cfg_attr]` attributes in its input ("fully configuring" the input).
The effect is identical to effect of `#[derive(Foo, Bar)]` which also fully configures its input before passing it to macros `Foo` and `Bar`, but unlike `#[derive]` `#[cfg_eval]` can be applied to any syntax nodes supporting macro attributes, not only certain items.
`cfg_eval` was the first name suggested in https://github.com/rust-lang/rust/pull/79078, but other alternatives are also possible, e.g. `cfg_expand`.
```rust
#[cfg_eval]
#[my_attr] // Receives `struct S {}` as input, the field is configured away by `#[cfg_eval]`
struct S {
#[cfg(FALSE)]
field: u8,
}
```
Tracking issue: https://github.com/rust-lang/rust/issues/82679
expand: Refactor module loading
This is an accompanying PR to https://github.com/rust-lang/rust/pull/82399, but they can be landed independently.
See individual commits for more details.
Anyone should be able to review this equally well because all people actually familiar with this code left the project.
Inherit `#[stable(..)]` annotations in enum variants and fields from its item
Lint changes for #65515. The stdlib will have to be updated once this lands in beta and that version is promoted in master.
When token-based attribute handling is implemeneted in #80689,
we will need to access tokens from `HasAttrs` (to perform
cfg-stripping), and we will to access attributes from `HasTokens` (to
construct a `PreexpTokenStream`).
This PR merges the `HasAttrs` and `HasTokens` traits into a new
`AstLike` trait. The previous `HasAttrs` impls from `Vec<Attribute>` and `AttrVec`
are removed - they aren't attribute targets, so the impls never really
made sense.
- Add the module name to `pre_AST_expansion_passes` and don't make it a
verbose event (since it normally doesn't take very long, and it's
emitted many times)
- Don't make the following rustdoc events verbose; they're emitted many times.
+ build_extern_trait_impl
+ build_local_trait_impl
+ build_primitive_trait_impl
+ get_auto_trait_impls
+ get_blanket_trait_impls
- Remove `get_auto_trait_and_blanket_synthetic_impls`; it's wholly
covered by get_{auto,blanket}_trait_impls and not very useful.
This makes it possible to have both std::panic and core::panic as a
builtin macro, by using different builtin macro names for each.
Also removes SyntaxExtension::is_derive_copy, as the macro name (e.g.
sym::Copy) is now tracked and provides that information directly.
