platform/upstream/gcc.git
19 months agogccrs: These are wrappers ported from reusing gccgo
Philip Herron [Wed, 24 Aug 2022 11:08:58 +0000 (12:08 +0100)]
gccrs: These are wrappers ported from reusing gccgo

The wrappers over linemap and location will eventually disappear here but
served as a useful starting point for us. We have wrappers over the
diagnostics system which we might be able to get rid of as well.

gcc/rust/
* rust-diagnostics.cc: New.
* rust-diagnostics.h: New.
* rust-gcc-diagnostics.cc: New.
* rust-linemap.cc: New.
* rust-linemap.h: New.
* rust-location.h: New.
* rust-system.h: New.

19 months agogccrs: Add HIR to GCC GENERIC lowering entry point
Philip Herron [Fri, 21 Oct 2022 12:29:50 +0000 (14:29 +0200)]
gccrs: Add HIR to GCC GENERIC lowering entry point

This patch contains the entry point and utilities used for the lowering
of HIR nodes to `tree`s. It also contains a constant evaluator, ported
over from the C++ frontend.

gcc/rust/
* backend/rust-compile-context.cc: New.
* backend/rust-compile-context.h: New.
* backend/rust-compile.cc: New.
* backend/rust-compile.h: New.
* backend/rust-constexpr.cc: New.
* backend/rust-constexpr.h: New.

Co-authored-by: David Faust <david.faust@oracle.com>
Co-authored-by: Faisal Abbas <90.abbasfaisal@gmail.com>
Signed-off-by: Faisal Abbas <90.abbasfaisal@gmail.com>
19 months agogccrs: Add HIR to GCC GENERIC lowering for all nodes
Philip Herron [Fri, 21 Oct 2022 12:27:56 +0000 (14:27 +0200)]
gccrs: Add HIR to GCC GENERIC lowering for all nodes

This patch implements the lowering mentioned in the previous patch for all HIR nodes.

gcc/rust/
* backend/rust-compile-block.cc: New.
* backend/rust-compile-block.h: New.
* backend/rust-compile-expr.cc: New.
* backend/rust-compile-expr.h: New.
* backend/rust-compile-extern.h: New.
* backend/rust-compile-fnparam.cc: New.
* backend/rust-compile-fnparam.h: New.
* backend/rust-compile-implitem.cc: New.
* backend/rust-compile-implitem.h: New.
* backend/rust-compile-intrinsic.cc: New.
* backend/rust-compile-intrinsic.h: New.
* backend/rust-compile-item.cc: New.
* backend/rust-compile-item.h: New.
* backend/rust-compile-pattern.cc: New.
* backend/rust-compile-pattern.h: New.
* backend/rust-compile-resolve-path.cc: New.
* backend/rust-compile-resolve-path.h: New.
* backend/rust-compile-stmt.cc: New.
* backend/rust-compile-stmt.h: New.
* backend/rust-compile-struct-field-expr.cc: New.
* backend/rust-compile-struct-field-expr.h: New.
* backend/rust-compile-type.cc: New.
* backend/rust-compile-type.h: New.
* backend/rust-compile-var-decl.h: New.

Co-authored-by: David Faust <david.faust@oracle.com>
19 months agogccrs: Add base for HIR to GCC GENERIC lowering
Philip Herron [Fri, 21 Oct 2022 12:01:04 +0000 (14:01 +0200)]
gccrs: Add base for HIR to GCC GENERIC lowering

This pass walks the HIR crate and turns them into GCC `tree`s. We do not have
any Rust specific tree's. We are slowly removing the backend abstraction
which was ported over from gccgo in favour of using `tree`s directly.

gcc/rust/
* backend/rust-builtins.h: New.
* backend/rust-compile-base.cc: New.
* backend/rust-compile-base.h: New.
* backend/rust-mangle.cc: New.
* backend/rust-mangle.h: New.
* backend/rust-tree.cc: New.
* backend/rust-tree.h: New.
* rust-backend.h: New.
* rust-gcc.cc: New.

Co-authored-by: David Faust <david.faust@oracle.com>
19 months agogccrs: Add metadata output pass
Philip Herron [Tue, 23 Aug 2022 15:36:29 +0000 (16:36 +0100)]
gccrs: Add metadata output pass

Extern crates statements to tell the front-end to look for another library.
The mechanism here is heavily inspired from gccgo, so when we compile a
library for example we invoke:

  gccrs -g -O2 -frust-crate=mylib -c src/lib.rs -o src/mylib.o

All going well this object file will now contain extra data inside
.rust-export section inside the object file which will be preserved inside
archives and shared objects. When we have another application which uses
this library 'mylib'.

  extern crate mylib;
  use mylib::foo;

  fn main() {
    foo();
  }

We compile using:

  gcc -g -O2 -frust-crate=test -c src/main.rs -o src/main.o

When the extern crate line is hit the front-end will look for mylib.o,
libmylib.a, mylib.rox. If it finds a raw object file it will read the
.rust-export section directly from the object for the public metadata
such as public functions, types constants etc. If it fails to find an
object it might find .rox which is the objdump of the .rust-export to a
raw file, it might even find libmylib.a and read the export directly out
of the archive file reusing code from gccgo to do so.

The full compiler pipeline is reused here, so the metatadata is actually
just real rust code. The benifit here is that Rust supports exporting,
macros and generics so this requires the name-resolution and type info
all to be generated and inserted into the apropriate context classes. Since
the metadata is real rust code it means we can reuse the full pipeline to
generate the code as necessary. So for the simple case of a public struct
we simply emit the AST dump of this struct directly into the metadata. If
its a non-generic public function we emit and extern rust abi block for
that function. If its a trait we can simply emit the trait with the public
memebers. Generics are more complicated since we need to emit the function
fully for it to be compiled correctly this still needs tests to be added.
The hardest part is non generic impl blocks which is still a WIP.

To finally link the two crates together you run:

  gcc -g -O2 -o rust-program.exe src/main.o src/mylib.o

gcc/rust/
* metadata/rust-export-metadata.cc: New.
* metadata/rust-export-metadata.h: New.
* metadata/rust-extern-crate.cc: New.
* metadata/rust-extern-crate.h: New.
* metadata/rust-import-archive.cc: New.
* metadata/rust-imports.cc: New.
* metadata/rust-imports.h: New.
* rust-object-export.cc: New.
* rust-object-export.h: New.

19 months agogccrs: Add unused variable scan
Philip Herron [Tue, 23 Aug 2022 15:35:44 +0000 (16:35 +0100)]
gccrs: Add unused variable scan

This is a simple walk_tree which acts on the monomorphized code by walking
the compiled translation unit of functions.

gcc/rust/
* checks/lints/rust-lint-unused-var.cc: New.
* checks/lints/rust-lint-unused-var.h: New.

19 months agogccrs: Add dead code scan on HIR
Thomas Young [Tue, 23 Aug 2022 15:34:35 +0000 (16:34 +0100)]
gccrs: Add dead code scan on HIR

In order to find dead code we use a depth first search and keep liveness
variables, after type resolution. In this case, if a function is unused
and it calls another function the 2nd function is now unused since the
caller is not used etc. The algorithm is a depth first search.

gcc/rust/
* checks/lints/rust-lint-marklive-base.h: New.
* checks/lints/rust-lint-marklive.cc: New.
* checks/lints/rust-lint-marklive.h: New.
* checks/lints/rust-lint-scan-deadcode.h: New.

19 months agogccrs: Add privacy checks
Arthur Cohen [Tue, 23 Aug 2022 15:32:26 +0000 (16:32 +0100)]
gccrs: Add privacy checks

This pass is responsible for resolving the privacy of items and verifying
that access to these items is performed within the limits of that privacy.
By default, items in Rust are private and only public to the current
module and its submodules. However, the user can annotate an item with
various qualifiers such as `pub` to publicly expose an item. Furthermore,
a module path can be given to `pub` to restrict an item's privacy to a
certain module: These paths need to be resolved and later on checked by
the privacy error reporter.

gcc/rust/
* checks/errors/privacy/rust-privacy-check.cc: New.
* checks/errors/privacy/rust-privacy-check.h: New.
* checks/errors/privacy/rust-privacy-common.h: New.
* checks/errors/privacy/rust-privacy-ctx.cc: New.
* checks/errors/privacy/rust-privacy-ctx.h: New.
* checks/errors/privacy/rust-privacy-reporter.cc: New.
* checks/errors/privacy/rust-privacy-reporter.h: New.
* checks/errors/privacy/rust-pub-restricted-visitor.cc: New.
* checks/errors/privacy/rust-pub-restricted-visitor.h: New.
* checks/errors/privacy/rust-reachability.cc: New.
* checks/errors/privacy/rust-reachability.h: New.
* checks/errors/privacy/rust-visibility-resolver.cc: New.
* checks/errors/privacy/rust-visibility-resolver.h: New.

19 months agogccrs: Add const checker
Arthur Cohen [Tue, 23 Aug 2022 15:31:55 +0000 (16:31 +0100)]
gccrs: Add const checker

Similarly to the unsafe checker, constant evaluation can only be performed
in a few contexts and include restrictions on the Rust language. Should
the user fail to uphold those conditions, errors will be reported and the
compilation pipeline interrupted.

These contexts are as follow:

- Array type length expressions
- Array repeat length expressions
- Constants
- Statics
- Enum discriminants
- Const generic arguments

In these contexts, the user is restricted to calling only functions marked
as `const` or perform arithmetic operations only on certain types, among
other restrictions.

gcc/rust/
* checks/errors/rust-const-checker.cc: New.
* checks/errors/rust-const-checker.h: New.

19 months agogccrs: Add unsafe checks for Rust
Arthur Cohen [Tue, 23 Aug 2022 15:31:17 +0000 (16:31 +0100)]
gccrs: Add unsafe checks for Rust

The UnsafeChecker visitor verifies that unsafe actions are only performed
in unsafe contexts. Otherwise, an error should be reported to the user and
the compilation pipeline should be halted. These contexts, which include
unsafe blocks or unsafe functions, are allowed to perform more actions
than regular safe Rust code. These actions currently include:

- Dereferencing raw pointers
- Calls to unsafe functions
- Use of inline assembly
- Use of mutable static
- Use of extern static
- Access to a union's field
- Call to functions with #[target(feature)] attribute
- Initializing type with rustc_layout_scalar_valid_range attribute
- Mutation of layout constrained field
- Borrow of layout constrained field

gcc/rust/
* checks/errors/rust-unsafe-checker.cc: New.
* checks/errors/rust-unsafe-checker.h: New.

19 months agogccrs: Add remaining type system transformations
Philip Herron [Fri, 21 Oct 2022 11:55:28 +0000 (13:55 +0200)]
gccrs: Add remaining type system transformations

This patch implements multiple transformation performed on the HIR
during type-resolution such as type coercion, casts, auto-dereferencement.

gcc/rust/
* typecheck/rust-autoderef.cc: New.
* typecheck/rust-autoderef.h: New.
* typecheck/rust-casts.cc: New.
* typecheck/rust-casts.h: New.
* typecheck/rust-coercion.cc: New.
* typecheck/rust-coercion.h: New.
* typecheck/rust-hir-dot-operator.cc: New.
* typecheck/rust-hir-dot-operator.h: New.
* typecheck/rust-hir-inherent-impl-overlap.h: New.
* typecheck/rust-hir-path-probe.h: New.
* typecheck/rust-hir-trait-ref.h: New.
* typecheck/rust-hir-type-bounds.h: New.
* typecheck/rust-substitution-mapper.cc: New.
* typecheck/rust-substitution-mapper.h: New.
* typecheck/rust-tycheck-dump.h: New.
* typecheck/rust-tyctx.cc: New.
* typecheck/rust-tyty-bounds.cc: New.
* typecheck/rust-tyty-call.cc: New.
* typecheck/rust-tyty-call.h: New.
* typecheck/rust-tyty-cmp.h: New.
* typecheck/rust-tyty-rules.h: New.

19 months agogccrs: Add Rust type information
Philip Herron [Fri, 21 Oct 2022 11:53:58 +0000 (13:53 +0200)]
gccrs: Add Rust type information

Contains abstractions over Rust's types, used when performing the
HIR's type-resolution.

gcc/rust/
* typecheck/rust-tyty.cc: New.
* typecheck/rust-tyty.h: New.

19 months agogccrs: Add type resolution and trait solving pass
Philip Herron [Fri, 21 Oct 2022 11:53:14 +0000 (13:53 +0200)]
gccrs: Add type resolution and trait solving pass

This serves to handle parts of the Rust type-system. Namely, the type
resolution (similar to type-checking) and the trait solving algorithms
(which ensure Rust's type contracts are upheld throughout the codebase).

gcc/rust/
* typecheck/rust-hir-trait-resolve.cc: New.
* typecheck/rust-hir-trait-resolve.h: New.
* typecheck/rust-hir-type-check-base.cc: New.
* typecheck/rust-hir-type-check-base.h: New.
* typecheck/rust-hir-type-check-enumitem.cc: New.
* typecheck/rust-hir-type-check-enumitem.h: New.
* typecheck/rust-hir-type-check-expr.cc: New.
* typecheck/rust-hir-type-check-expr.h: New.
* typecheck/rust-hir-type-check-implitem.cc: New.
* typecheck/rust-hir-type-check-implitem.h: New.
* typecheck/rust-hir-type-check-item.cc: New.
* typecheck/rust-hir-type-check-item.h: New.
* typecheck/rust-hir-type-check-path.cc: New.
* typecheck/rust-hir-type-check-pattern.cc: New.
* typecheck/rust-hir-type-check-pattern.h: New.
* typecheck/rust-hir-type-check-stmt.cc: New.
* typecheck/rust-hir-type-check-stmt.h: New.
* typecheck/rust-hir-type-check-struct-field.h: New.
* typecheck/rust-hir-type-check-struct.cc: New.
* typecheck/rust-hir-type-check-toplevel.cc: New.
* typecheck/rust-hir-type-check-toplevel.h: New.
* typecheck/rust-hir-type-check-type.cc: New.
* typecheck/rust-hir-type-check-type.h: New.
* typecheck/rust-hir-type-check-util.cc: New.
* typecheck/rust-hir-type-check-util.h: New.
* typecheck/rust-hir-type-check.cc: New.
* typecheck/rust-hir-type-check.h: New.
* typecheck/rust-tyty-visitor.h: New.

19 months agogccrs: Add helpers mappings canonical path and lang items
Philip Herron [Tue, 23 Aug 2022 15:29:11 +0000 (16:29 +0100)]
gccrs: Add helpers mappings canonical path and lang items

These are various helper classes used in the compiler pipeline.

gcc/rust/
* util/rust-canonical-path.h: New.
* util/rust-common.h: New.
* util/rust-hir-map.cc: New.
* util/rust-hir-map.h: New.
* util/rust-identifier.h: New.
* util/rust-lang-item.h: New.
* util/rust-mapping-common.h: New.
* util/rust-stacked-contexts.h: New.

19 months agogccrs: Add attributes checker
Arthur Cohen [Tue, 23 Aug 2022 15:27:23 +0000 (16:27 +0100)]
gccrs: Add attributes checker

The attribute checker is responsible for checking the validity of various
attributes including built-in ones. It is currently unfinished and will
receive some modifications, as well as become the host of some existing
code in the compiler which needs to be refactored. One of its
responsibilities is to make sure that arguments given to built-in
attributes are correct, or contain the correct type of information. This
visitor also checks that an attribute is allowed to be used in the current
particular context.

gcc/rust/
* util/rust-attributes.cc: New.
* util/rust-attributes.h: New.

19 months agogccrs: Add implementation of Optional
Arthur Cohen [Tue, 23 Aug 2022 15:26:37 +0000 (16:26 +0100)]
gccrs: Add implementation of Optional

This adds an Optional<T> class to improve error handling.

gcc/rust/
* util/rust-optional-test.cc: New.
* util/rust-optional.h: New.

19 months agogccrs: Add Base62 implementation
Arthur Cohen [Tue, 23 Aug 2022 15:26:01 +0000 (16:26 +0100)]
gccrs: Add Base62 implementation

Used for V0 symbol mangling scheme which is still in development.

gcc/rust/
* util/rust-base62.cc: New.
* util/rust-base62.h: New.

19 months agogccrs: Add Rust ABI enum helpers
Philip Herron [Tue, 23 Aug 2022 15:23:22 +0000 (16:23 +0100)]
gccrs: Add Rust ABI enum helpers

This is a simple helper over an enum of possible ABI options in Rust.

gcc/rust/
* util/rust-abi.cc: New.
* util/rust-abi.h: New.

19 months agogccrs: Add port of FNV hash used during legacy symbol mangling
Philip Herron [Tue, 23 Aug 2022 15:22:47 +0000 (16:22 +0100)]
gccrs: Add port of FNV hash used during legacy symbol mangling

This hash was ported from the Go runtime, as we needed a hash for the legacy
symbol mangling system. All symbols in Rust contain a hash of some
metadata for uniqueness on generic functions.

gcc/rust/
* util/fnv-hash.h: New.

19 months agogccrs: Add wrapper for make_unique
Philip Herron [Tue, 23 Aug 2022 15:22:22 +0000 (16:22 +0100)]
gccrs: Add wrapper for make_unique

This is a wrapper for make_unique. We can likely get rid of this, as there
are other implementations available, or simply keep using the unique_ptr
constructor.

gcc/rust/
* util/rust-make-unique.h: New.

19 months agogccrs: Add AST to HIR lowering pass
Philip Herron [Tue, 23 Aug 2022 15:20:05 +0000 (16:20 +0100)]
gccrs: Add AST to HIR lowering pass

This performs the lowering of the AST to HIR. The interesting piece here is
that we desugar much of the AST as mentioned in the previous pass, but
crucially, we also strip out all code that is "marked-for-strip" which failed
cfg-expansion from the expansion pass. After this, the HIR includes all code
required to compile for this crate.

gcc/rust/
* hir/rust-ast-lower-base.cc: New.
* hir/rust-ast-lower-base.h: New.
* hir/rust-ast-lower-block.h: New.
* hir/rust-ast-lower-enumitem.h: New.
* hir/rust-ast-lower-expr.h: New.
* hir/rust-ast-lower-extern.h: New.
* hir/rust-ast-lower-implitem.h: New.
* hir/rust-ast-lower-item.cc: New.
* hir/rust-ast-lower-item.h: New.
* hir/rust-ast-lower-pattern.cc: New.
* hir/rust-ast-lower-pattern.h: New.
* hir/rust-ast-lower-stmt.h: New.
* hir/rust-ast-lower-struct-field-expr.h: New.
* hir/rust-ast-lower-type.h: New.
* hir/rust-ast-lower.cc: New.
* hir/rust-ast-lower.h: New.
* hir/rust-hir-dump.cc: New.
* hir/rust-hir-dump.h: New.

19 months agogccrs: Add HIR definitions and visitor framework
Philip Herron [Fri, 21 Oct 2022 11:45:38 +0000 (13:45 +0200)]
gccrs: Add HIR definitions and visitor framework

This patch implements the classes mentioned in the previous HIR patch,
as well as a set of visitor frameworks used in handling that HIR.

gcc/rust/
* hir/tree/rust-hir-full-decls.h: New.
* hir/tree/rust-hir-full-test.cc: New.
* hir/tree/rust-hir-full.h: New.
* hir/tree/rust-hir-visitor.h: New.
* hir/tree/rust-hir.h: New.

19 months agogccrs: Add declarations for Rust HIR
Philip Herron [Fri, 21 Oct 2022 11:44:20 +0000 (13:44 +0200)]
gccrs: Add declarations for Rust HIR

This patch contains the declarations needed for our second intermediate
representation, which we will refer to as an HIR.

This gives the front-end a chance to desugar much of the AST, such as:
- Removing distinction between functions and methods
- Removing Macros
- Removing IdentifierExprs
- Removing duplicate attribute structures

gcc/rust/
* hir/tree/rust-hir-expr.h: New.
* hir/tree/rust-hir-item.h: New.
* hir/tree/rust-hir-path.h: New.
* hir/tree/rust-hir-pattern.h: New.
* hir/tree/rust-hir-stmt.h: New.
* hir/tree/rust-hir-type.h: New.

19 months agogccrs: Add name resolution pass to the Rust front-end
Philip Herron [Tue, 23 Aug 2022 15:19:04 +0000 (16:19 +0100)]
gccrs: Add name resolution pass to the Rust front-end

The name resolution is split into two phases, one toplevel pass which scans
the whole "Crate" which iterates all items and nested items in modules to
generate a context class full of CanonicalPath items. It also generates
a hierarchy of parent->child and child->parent relationships using the AST
NodeId for PathResolution in the second phase.

The second phase drills into each item like functions and creates a stack
of canonical paths for variables etc so that we can store information in
a side table of usage variable 'a' resolves to NodeId '123' which refers
to the NodeId of the "let a;" statement.

gcc/rust/
* resolve/rust-ast-resolve-base.cc: New.
* resolve/rust-ast-resolve-base.h: New.
* resolve/rust-ast-resolve-expr.cc: New.
* resolve/rust-ast-resolve-expr.h: New.
* resolve/rust-ast-resolve-implitem.h: New.
* resolve/rust-ast-resolve-item.cc: New.
* resolve/rust-ast-resolve-item.h: New.
* resolve/rust-ast-resolve-path.cc: New.
* resolve/rust-ast-resolve-path.h: New.
* resolve/rust-ast-resolve-pattern.cc: New.
* resolve/rust-ast-resolve-pattern.h: New.
* resolve/rust-ast-resolve-stmt.cc: New.
* resolve/rust-ast-resolve-stmt.h: New.
* resolve/rust-ast-resolve-struct-expr-field.cc: New.
* resolve/rust-ast-resolve-struct-expr-field.h: New.
* resolve/rust-ast-resolve-toplevel.h: New.
* resolve/rust-ast-resolve-type.cc: New.
* resolve/rust-ast-resolve-type.h: New.
* resolve/rust-ast-resolve.cc: New.
* resolve/rust-ast-resolve.h: New.
* resolve/rust-ast-verify-assignee.h: New.
* resolve/rust-name-resolver.cc: New.
* resolve/rust-name-resolver.h: New.

19 months agogccrs: Add expansion pass for the Rust front-end
Arthur Cohen [Tue, 23 Aug 2022 15:15:22 +0000 (16:15 +0100)]
gccrs: Add expansion pass for the Rust front-end

The expansion pass is responsible for two actions on our AST:

1. Expanding macro calls
2. Performing conditional compilation

Calls to macros should be checked and expanded into an AST fragment based on
the context they've been called in. This is similar to token substitution, with
a lot of intricacies and checks being performed. A single invocation can result
in an AST fragment containing multiple statements or multiple expressions,
which need to be handled as well. Furthermore, Rust macros can contain
repetitions relying on Kleine operators, similar to regular expression
patterns, that also need to be expanded properly.

Finally, Rust code can be hidden behind `cfg` directives, which allow the user
to perform conditional compilation. If a `cfg` predicate is not met, the
expression or statement it refers to should be marked for strip and removed
from the AST.

gcc/rust/
* expand/rust-attribute-visitor.cc: New.
* expand/rust-attribute-visitor.h: New.
* expand/rust-macro-builtins.cc: New.
* expand/rust-macro-builtins.h: New.
* expand/rust-macro-expand.cc: New.
* expand/rust-macro-expand.h: New.
* expand/rust-macro-invoc-lexer.cc: New.
* expand/rust-macro-invoc-lexer.h: New.
* expand/rust-macro-substitute-ctx.cc: New.
* expand/rust-macro-substitute-ctx.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Co-authored-by: Joel Phillips <simplytheother@gmail.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add Parser for Rust front-end pt.2
Joel Phillips [Fri, 21 Oct 2022 11:39:49 +0000 (13:39 +0200)]
gccrs: Add Parser for Rust front-end pt.2

This patch contains the second half of the templated Rust parser.

gcc/rust/
* parse/rust-parse-impl.h: New, second half.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add Parser for Rust front-end pt.1
Joel Phillips [Fri, 21 Oct 2022 11:39:04 +0000 (13:39 +0200)]
gccrs: Add Parser for Rust front-end pt.1

This is a Pratt-style parser for Rust that implements all of the AST. The
rust-parser-impl.h is the implementation of the parser as a template,
allowing it to be given ManagedTokenSource and avoid virtual calls. The
downside is it takes time to compile when used.

see: https://en.wikipedia.org/wiki/Operator-precedence_parser#Pratt_parsing

This patch contains the first half of the templated parser, so as to not
lose patches in the mailing list archives.

gcc/rust/
* parse/rust-cfg-parser.cc: New.
* parse/rust-cfg-parser.h: New.
* parse/rust-parse-impl.h: New.
* parse/rust-parse.cc: New.
* parse/rust-parse.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add Lexer for Rust front-end
Joel Phillips [Tue, 23 Aug 2022 15:11:00 +0000 (16:11 +0100)]
gccrs: Add Lexer for Rust front-end

The lexer is referred to as a ManagedTokenSource within the parser. This
lexer does not currently support Unicode, but serves as a starting point
to do so.

gcc/rust/
* lex/rust-codepoint.h: New.
* lex/rust-lex.cc: New.
* lex/rust-lex.h: New.
* lex/rust-token.cc: New.
* lex/rust-token.h: New.
* rust-buffered-queue.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Co-authored-by: Mark Wielaard <mark@klomp.org>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add Rust AST visitors
Joel Phillips [Fri, 21 Oct 2022 11:35:58 +0000 (13:35 +0200)]
gccrs: Add Rust AST visitors

This patch contains the basic framework of our AST visitors, as well as
one aimed at pretty-printing and exporting these AST nodes.

gcc/rust/
* ast/rust-ast-dump.cc: New.
* ast/rust-ast-dump.h: New.
* ast/rust-ast-visitor.h: New.
* ast/rust-cond-compilation.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add full definitions of Rust AST data structures
Joel Phillips [Fri, 21 Oct 2022 12:35:54 +0000 (14:35 +0200)]
gccrs: Add full definitions of Rust AST data structures

This adds the proper definitions of our AST nodes split across multiple
files for clarity

gcc/rust/
* ast/rust-expr.h: New.
* ast/rust-macro.h: New.
* ast/rust-path.h: New.
* ast/rust-pattern.h: New.
* ast/rust-stmt.h: New.
* ast/rust-type.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add definitions of Rust Items in AST data structures
Joel Phillips [Fri, 21 Oct 2022 12:35:09 +0000 (14:35 +0200)]
gccrs: Add definitions of Rust Items in AST data structures

This adds the proper definitions of our AST Item nodes.

gcc/rust/
* ast/rust-item.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add Rust front-end base AST data structures
Joel Phillips [Fri, 21 Oct 2022 11:34:11 +0000 (13:34 +0200)]
gccrs: Add Rust front-end base AST data structures

This is a full C++11 class hierarchy representing the Rust AST. We do not
allow dynamic_cast and so the main mechanism to work with the AST is by
using the visitor interface. Slowly we are adding TREE_CODE style node
types to the AST which will allow for more ways to work with the AST but
for now this is it.

See: https://doc.rust-lang.org/reference/items.html

gcc/rust/
* ast/rust-ast-full-decls.h: New.
* ast/rust-ast-full-test.cc: New.
* ast/rust-ast-full.h: New.
* ast/rust-ast.h: New.
* operator.h: New.

Co-authored-by: Philip Herron <philip.herron@embecosm.com>
Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Signed-off-by: Joel Phillips <simplytheother@gmail.com>
19 months agogccrs: Add gcc-check-target check-rust
Philip Herron [Tue, 23 Aug 2022 15:04:30 +0000 (16:04 +0100)]
gccrs: Add gcc-check-target check-rust

This allows us to invoke the rust testsuite.

* Makefile.def: Add Rust language.
* Makefile.in: Regenerate via autogen.

19 months agogccrs: Add execution test cases
Philip Herron [Tue, 23 Aug 2022 15:02:25 +0000 (16:02 +0100)]
gccrs: Add execution test cases

This is similar to the compile/torture/*.rs test cases, but all of these are
dg-execute testcases. They are compiled, linked and executed by default. These
testcases are also compiled with the matrix of torture options.

The only caveat here is that gccrs does not currently support the main shim,
so we have a C-style main function here returning zero which is not proper Rust
code.

gcc/testsuite/
* rust/execute/torture/block_expr1.rs: New.
* rust/execute/torture/builtin_macro_cfg.rs: New.
* rust/execute/torture/builtin_macro_concat.rs: New.
* rust/execute/torture/builtin_macro_env.rs: New.
* rust/execute/torture/builtin_macro_include_bytes.rs: New.
* rust/execute/torture/builtin_macro_include_str.rs: New.
* rust/execute/torture/builtin_macro_line.rs: New.
* rust/execute/torture/builtin_macros1.rs: New.
* rust/execute/torture/builtin_macros3.rs: New.
* rust/execute/torture/cfg1.rs: New.
* rust/execute/torture/cfg2.rs: New.
* rust/execute/torture/cfg3.rs: New.
* rust/execute/torture/cfg4.rs: New.
* rust/execute/torture/cfg5.rs: New.
* rust/execute/torture/coercion1.rs: New.
* rust/execute/torture/coercion2.rs: New.
* rust/execute/torture/const_fold1.rs: New.
* rust/execute/torture/const_fold2.rs: New.
* rust/execute/torture/copy_nonoverlapping1.rs: New.
* rust/execute/torture/empty_main.rs: New.
* rust/execute/torture/execute.exp: New.
* rust/execute/torture/exit_error.rs: New.
* rust/execute/torture/extern_mod4.rs: New.
* rust/execute/torture/func1.rs: New.
* rust/execute/torture/helloworld1.rs: New.
* rust/execute/torture/helloworld2.rs: New.
* rust/execute/torture/include.txt: New.
* rust/execute/torture/index1.rs: New.
* rust/execute/torture/issue-1120.rs: New.
* rust/execute/torture/issue-1133.rs: New.
* rust/execute/torture/issue-1198.rs: New.
* rust/execute/torture/issue-1231.rs: New.
* rust/execute/torture/issue-1232.rs: New.
* rust/execute/torture/issue-1249.rs: New.
* rust/execute/torture/issue-1436.rs: New.
* rust/execute/torture/issue-1496.rs: New.
* rust/execute/torture/issue-647.rs: New.
* rust/execute/torture/issue-845.rs: New.
* rust/execute/torture/issue-851.rs: New.
* rust/execute/torture/issue-858.rs: New.
* rust/execute/torture/issue-976.rs: New.
* rust/execute/torture/issue-995.rs: New.
* rust/execute/torture/macros1.rs: New.
* rust/execute/torture/macros10.rs: New.
* rust/execute/torture/macros11.rs: New.
* rust/execute/torture/macros12.rs: New.
* rust/execute/torture/macros13.rs: New.
* rust/execute/torture/macros14.rs: New.
* rust/execute/torture/macros16.rs: New.
* rust/execute/torture/macros17.rs: New.
* rust/execute/torture/macros18.rs: New.
* rust/execute/torture/macros19.rs: New.
* rust/execute/torture/macros2.rs: New.
* rust/execute/torture/macros20.rs: New.
* rust/execute/torture/macros21.rs: New.
* rust/execute/torture/macros22.rs: New.
* rust/execute/torture/macros23.rs: New.
* rust/execute/torture/macros24.rs: New.
* rust/execute/torture/macros25.rs: New.
* rust/execute/torture/macros26.rs: New.
* rust/execute/torture/macros27.rs: New.
* rust/execute/torture/macros28.rs: New.
* rust/execute/torture/macros29.rs: New.
* rust/execute/torture/macros3.rs: New.
* rust/execute/torture/macros30.rs: New.
* rust/execute/torture/macros31.rs: New.
* rust/execute/torture/macros4.rs: New.
* rust/execute/torture/macros5.rs: New.
* rust/execute/torture/macros6.rs: New.
* rust/execute/torture/macros7.rs: New.
* rust/execute/torture/macros8.rs: New.
* rust/execute/torture/macros9.rs: New.
* rust/execute/torture/match1.rs: New.
* rust/execute/torture/match2.rs: New.
* rust/execute/torture/match3.rs: New.
* rust/execute/torture/match_bool1.rs: New.
* rust/execute/torture/match_byte1.rs: New.
* rust/execute/torture/match_char1.rs: New.
* rust/execute/torture/match_int1.rs: New.
* rust/execute/torture/match_loop1.rs: New.
* rust/execute/torture/match_range1.rs: New.
* rust/execute/torture/match_range2.rs: New.
* rust/execute/torture/match_tuple1.rs: New.
* rust/execute/torture/method1.rs: New.
* rust/execute/torture/method2.rs: New.
* rust/execute/torture/method3.rs: New.
* rust/execute/torture/method4.rs: New.
* rust/execute/torture/mod1.rs: New.
* rust/execute/torture/modules/mod.rs: New.
* rust/execute/torture/operator_overload_1.rs: New.
* rust/execute/torture/operator_overload_10.rs: New.
* rust/execute/torture/operator_overload_11.rs: New.
* rust/execute/torture/operator_overload_12.rs: New.
* rust/execute/torture/operator_overload_2.rs: New.
* rust/execute/torture/operator_overload_3.rs: New.
* rust/execute/torture/operator_overload_4.rs: New.
* rust/execute/torture/operator_overload_5.rs: New.
* rust/execute/torture/operator_overload_6.rs: New.
* rust/execute/torture/operator_overload_7.rs: New.
* rust/execute/torture/operator_overload_8.rs: New.
* rust/execute/torture/operator_overload_9.rs: New.
* rust/execute/torture/slice-magic.rs: New.
* rust/execute/torture/slice-magic2.rs: New.
* rust/execute/torture/slice1.rs: New.
* rust/execute/torture/str-layout1.rs: New.
* rust/execute/torture/str-zero.rs: New.
* rust/execute/torture/trait1.rs: New.
* rust/execute/torture/trait10.rs: New.
* rust/execute/torture/trait11.rs: New.
* rust/execute/torture/trait12.rs: New.
* rust/execute/torture/trait13.rs: New.
* rust/execute/torture/trait2.rs: New.
* rust/execute/torture/trait3.rs: New.
* rust/execute/torture/trait4.rs: New.
* rust/execute/torture/trait5.rs: New.
* rust/execute/torture/trait6.rs: New.
* rust/execute/torture/trait7.rs: New.
* rust/execute/torture/trait8.rs: New.
* rust/execute/torture/trait9.rs: New.
* rust/execute/torture/transmute1.rs: New.
* rust/execute/torture/wrapping_op1.rs: New.
* rust/execute/torture/wrapping_op2.rs: New.
* rust/execute/xfail/macro1.rs: New.

Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Co-authored-by: Thomas Schwinge <thomas@codesourcery.com>
Co-authored-by: Mark Wielaard <mark@klomp.org>
Co-authored-by: Marc Poulhiès <dkm@kataplop.net>
19 months agogccrs: Add general compilation test cases
Philip Herron [Tue, 23 Aug 2022 15:01:16 +0000 (16:01 +0100)]
gccrs: Add general compilation test cases

This suite of tests has two sections: compile/*.rs and compile/torture/*.rs.
The first section is comprised entirely of dg-compile tests, containing a
mixture of dg-warning and dg-error annotations and some with no annotations,
ensuring the creation of resulting asm output. The second section is the same,
but has tests which are ran with the full torture options, for coverage of test
cases that may have an issue with a specific optimization level.

gcc/testsuite/
* rust/compile/abi-options1.rs: New.
* rust/compile/array3.rs: New.
* rust/compile/array_empty_list.rs: New.
* rust/compile/arrays1.rs: New.
* rust/compile/arrays2.rs: New.
* rust/compile/attr-mismatch-crate-name.rs: New.
* rust/compile/attr_cold.rs: New.
* rust/compile/attr_deprecated.rs: New.
* rust/compile/attr_deprecated_2.rs: New.
* rust/compile/bad-crate-name.rs: New.
* rust/compile/bad=file-name.rs: New.
* rust/compile/bad_as_bool_char.rs: New.
* rust/compile/bad_file_name.txt.rs: New.
* rust/compile/bad_inner_doc.rs: New.
* rust/compile/bad_pub_enumitems.rs: New.
* rust/compile/bad_stmt_enums.rs: New.
* rust/compile/bad_toplevel_enums.rs: New.
* rust/compile/bad_tuple_index.rs: New.
* rust/compile/bad_type1.rs: New.
* rust/compile/bad_type2.rs: New.
* rust/compile/break1.rs: New.
* rust/compile/break2.rs: New.
* rust/compile/builtin_macro_compile_error.rs: New.
* rust/compile/builtin_macro_concat.rs: New.
* rust/compile/builtin_macro_env.rs: New.
* rust/compile/builtin_macro_include_bytes.rs: New.
* rust/compile/builtin_macro_include_str.rs: New.
* rust/compile/builtin_macro_not_found.rs: New.
* rust/compile/bytecharstring.rs: New.
* rust/compile/canonical_paths1.rs: New.
* rust/compile/cast1.rs: New.
* rust/compile/cfg1.rs: New.
* rust/compile/cfg2.rs: New.
* rust/compile/cfg3.rs: New.
* rust/compile/cfg4.rs: New.
* rust/compile/cfg5.rs: New.
* rust/compile/compile.exp: New.
* rust/compile/complex-path1.rs: New.
* rust/compile/const-issue1440.rs: New.
* rust/compile/const1.rs: New.
* rust/compile/const2.rs: New.
* rust/compile/const3.rs: New.
* rust/compile/const_generics_1.rs: New.
* rust/compile/const_generics_2.rs: New.
* rust/compile/const_generics_3.rs: New.
* rust/compile/const_generics_4.rs: New.
* rust/compile/const_generics_5.rs: New.
* rust/compile/const_generics_6.rs: New.
* rust/compile/continue1.rs: New.
* rust/compile/deadcode_err1.rs: New.
* rust/compile/deadcode_err2.rs: New.
* rust/compile/debug-diagnostics-default.rs: New.
* rust/compile/debug-diagnostics-off.rs: New.
* rust/compile/debug-diagnostics-on.rs: New.
* rust/compile/doc_isolated_cr_block_comment.rs: New.
* rust/compile/doc_isolated_cr_inner_block_comment.rs: New.
* rust/compile/doc_isolated_cr_inner_line_comment.rs: New.
* rust/compile/doc_isolated_cr_line_comment.rs: New.
* rust/compile/dup_fields.rs: New.
* rust/compile/empty_comment_before_match.rs: New.
* rust/compile/expected_type_args2.rs: New.
* rust/compile/expected_type_args3.rs: New.
* rust/compile/func1.rs: New.
* rust/compile/func2.rs: New.
* rust/compile/func3.rs: New.
* rust/compile/func4.rs: New.
* rust/compile/func5.rs: New.
* rust/compile/generic-default1.rs: New.
* rust/compile/generics1.rs: New.
* rust/compile/generics10.rs: New.
* rust/compile/generics11.rs: New.
* rust/compile/generics12.rs: New.
* rust/compile/generics13.rs: New.
* rust/compile/generics2.rs: New.
* rust/compile/generics3.rs: New.
* rust/compile/generics4.rs: New.
* rust/compile/generics5.rs: New.
* rust/compile/generics6.rs: New.
* rust/compile/generics7.rs: New.
* rust/compile/generics8.rs: New.
* rust/compile/generics9.rs: New.
* rust/compile/implicit_returns_err1.rs: New.
* rust/compile/implicit_returns_err2.rs: New.
* rust/compile/implicit_returns_err3.rs: New.
* rust/compile/implicit_returns_err4.rs: New.
* rust/compile/infer-crate-name.rs: New.
* rust/compile/inline_1.rs: New.
* rust/compile/inline_2.rs: New.
* rust/compile/issue-1005.rs: New.
* rust/compile/issue-1019.rs: New.
* rust/compile/issue-1023.rs: New.
* rust/compile/issue-1031.rs: New.
* rust/compile/issue-1034.rs: New.
* rust/compile/issue-1089.rs: New.
* rust/compile/issue-1128.rs: New.
* rust/compile/issue-1129-1.rs: New.
* rust/compile/issue-1129-2.rs: New.
* rust/compile/issue-1130.rs: New.
* rust/compile/issue-1131.rs: New.
* rust/compile/issue-1152.rs: New.
* rust/compile/issue-1165.rs: New.
* rust/compile/issue-1173.rs: New.
* rust/compile/issue-1226.rs: New.
* rust/compile/issue-1234.rs: New.
* rust/compile/issue-1235.rs: New.
* rust/compile/issue-1237.rs: New.
* rust/compile/issue-1251.rs: New.
* rust/compile/issue-1271.rs: New.
* rust/compile/issue-1289.rs: New.
* rust/compile/issue-1323-1.rs: New.
* rust/compile/issue-1323-2.rs: New.
* rust/compile/issue-1383.rs: New.
* rust/compile/issue-1393.rs: New.
* rust/compile/issue-1447.rs: New.
* rust/compile/issue-407-2.rs: New.
* rust/compile/issue-407.rs: New.
* rust/compile/issue-557.rs: New.
* rust/compile/issue-635-1.rs: New.
* rust/compile/issue-635-2.rs: New.
* rust/compile/lookup_err1.rs: New.
* rust/compile/macro-issue1053-2.rs: New.
* rust/compile/macro-issue1053.rs: New.
* rust/compile/macro-issue1224.rs: New.
* rust/compile/macro-issue1233.rs: New.
* rust/compile/macro-issue1395-2.rs: New.
* rust/compile/macro-issue1395.rs: New.
* rust/compile/macro-issue1400-2.rs: New.
* rust/compile/macro-issue1400.rs: New.
* rust/compile/macro1.rs: New.
* rust/compile/macro10.rs: New.
* rust/compile/macro11.rs: New.
* rust/compile/macro12.rs: New.
* rust/compile/macro13.rs: New.
* rust/compile/macro14.rs: New.
* rust/compile/macro15.rs: New.
* rust/compile/macro16.rs: New.
* rust/compile/macro17.rs: New.
* rust/compile/macro18.rs: New.
* rust/compile/macro19.rs: New.
* rust/compile/macro2.rs: New.
* rust/compile/macro20.rs: New.
* rust/compile/macro21.rs: New.
* rust/compile/macro22.rs: New.
* rust/compile/macro23.rs: New.
* rust/compile/macro25.rs: New.
* rust/compile/macro26.rs: New.
* rust/compile/macro27.rs: New.
* rust/compile/macro28.rs: New.
* rust/compile/macro29.rs: New.
* rust/compile/macro3.rs: New.
* rust/compile/macro30.rs: New.
* rust/compile/macro31.rs: New.
* rust/compile/macro32.rs: New.
* rust/compile/macro33.rs: New.
* rust/compile/macro34.rs: New.
* rust/compile/macro35.rs: New.
* rust/compile/macro36.rs: New.
* rust/compile/macro37.rs: New.
* rust/compile/macro38.rs: New.
* rust/compile/macro39.rs: New.
* rust/compile/macro4.rs: New.
* rust/compile/macro40.rs: New.
* rust/compile/macro41.rs: New.
* rust/compile/macro42.rs: New.
* rust/compile/macro5.rs: New.
* rust/compile/macro6.rs: New.
* rust/compile/macro7.rs: New.
* rust/compile/macro8.rs: New.
* rust/compile/macro9.rs: New.
* rust/compile/macro_return.rs: New.
* rust/compile/match1.rs: New.
* rust/compile/match2.rs: New.
* rust/compile/match3.rs: New.
* rust/compile/match4.rs: New.
* rust/compile/match5.rs: New.
* rust/compile/match6.rs: New.
* rust/compile/match7.rs: New.
* rust/compile/method1.rs: New.
* rust/compile/method2.rs: New.
* rust/compile/mismatch-crate-name.rs: New.
* rust/compile/missing_middle/both_path.rs: New.
* rust/compile/missing_middle/explicit.not.rs: New.
* rust/compile/missing_middle/inner_path.rs: New.
* rust/compile/missing_middle/other.rs: New.
* rust/compile/missing_middle/outer_path.rs: New.
* rust/compile/missing_middle/sub/mod.rs: New.
* rust/compile/missing_return1.rs: New.
* rust/compile/mod_missing_middle.rs: New.
* rust/compile/never_type_err1.rs: New.
* rust/compile/privacy1.rs: New.
* rust/compile/privacy2.rs: New.
* rust/compile/privacy3.rs: New.
* rust/compile/privacy4.rs: New.
* rust/compile/privacy5.rs: New.
* rust/compile/privacy6.rs: New.
* rust/compile/pub_restricted_1.rs: New.
* rust/compile/pub_restricted_2.rs: New.
* rust/compile/pub_restricted_3.rs: New.
* rust/compile/raw_identifiers_bad_keywords.rs: New.
* rust/compile/raw_identifiers_underscore.rs: New.
* rust/compile/redef_error1.rs: New.
* rust/compile/redef_error2.rs: New.
* rust/compile/redef_error3.rs: New.
* rust/compile/redef_error4.rs: New.
* rust/compile/redef_error5.rs: New.
* rust/compile/redef_error6.rs: New.
* rust/compile/reference1.rs: New.
* rust/compile/self-path1.rs: New.
* rust/compile/self-path2.rs: New.
* rust/compile/shadow1.rs: New.
* rust/compile/specify-crate-name.rs: New.
* rust/compile/static_var1.rs: New.
* rust/compile/stmt_with_block_err1.rs: New.
* rust/compile/struct_align1.rs: New.
* rust/compile/struct_align2.rs: New.
* rust/compile/struct_init1.rs: New.
* rust/compile/struct_pack1.rs: New.
* rust/compile/struct_pack2.rs: New.
* rust/compile/syntax-only.rs: New.
* rust/compile/test_mod.rs: New.
* rust/compile/torture/all_doc_comment_line_blocks.rs: New.
* rust/compile/torture/all_doc_comment_line_blocks_crlf.rs: New.
* rust/compile/torture/arithmetic_expressions1.rs: New.
* rust/compile/torture/array_const_fold_1.rs: New.
* rust/compile/torture/array_const_fold_2.rs: New.
* rust/compile/torture/array_function.rs: New.
* rust/compile/torture/array_type_infer.rs: New.
* rust/compile/torture/array_zero_length.rs: New.
* rust/compile/torture/arrays1.rs: New.
* rust/compile/torture/arrays2.rs: New.
* rust/compile/torture/arrays3.rs: New.
* rust/compile/torture/arrays4.rs: New.
* rust/compile/torture/arrays5.rs: New.
* rust/compile/torture/arrays6.rs: New.
* rust/compile/torture/arrays_index1.rs: New.
* rust/compile/torture/arrays_index2.rs: New.
* rust/compile/torture/arrays_index3.rs: New.
* rust/compile/torture/as_bool_char.rs: New.
* rust/compile/torture/associated_types1.rs: New.
* rust/compile/torture/autoderef1.rs: New.
* rust/compile/torture/block_expr1.rs: New.
* rust/compile/torture/block_expr2.rs: New.
* rust/compile/torture/block_expr3.rs: New.
* rust/compile/torture/block_expr4.rs: New.
* rust/compile/torture/block_expr5.rs: New.
* rust/compile/torture/block_expr_parser_bug.rs: New.
* rust/compile/torture/bom.rs: New.
* rust/compile/torture/bom_comment.rs: New.
* rust/compile/torture/bom_shebang.rs: New.
* rust/compile/torture/bom_whitespace.rs: New.
* rust/compile/torture/bools_eq.rs: New.
* rust/compile/torture/borrow1.rs: New.
* rust/compile/torture/borrow_function.rs: New.
* rust/compile/torture/break_function.rs: New.
* rust/compile/torture/byte_char_str.rs: New.
* rust/compile/torture/byte_str.rs: New.
* rust/compile/torture/cast1.rs: New.
* rust/compile/torture/cast2.rs: New.
* rust/compile/torture/cast3.rs: New.
* rust/compile/torture/cfg_attr.rs: New.
* rust/compile/torture/char1.rs: New.
* rust/compile/torture/check-doc-attr-string.rs: New.
* rust/compile/torture/coercion1.rs: New.
* rust/compile/torture/coercion2.rs: New.
* rust/compile/torture/comparison_expr1.rs: New.
* rust/compile/torture/compile.exp: New.
* rust/compile/torture/compound_assignment_expr1.rs: New.
* rust/compile/torture/conditional.rs: New.
* rust/compile/torture/constant1.rs: New.
* rust/compile/torture/constant2.rs: New.
* rust/compile/torture/constant3.rs: New.
* rust/compile/torture/deadcode1.rs: New.
* rust/compile/torture/deadcode2.rs: New.
* rust/compile/torture/deref1.rs: New.
* rust/compile/torture/deref_function.rs: New.
* rust/compile/torture/doc_comment.rs: New.
* rust/compile/torture/enum1.rs: New.
* rust/compile/torture/extern_mod1.rs: New.
* rust/compile/torture/extern_mod2.rs: New.
* rust/compile/torture/float1.rs: New.
* rust/compile/torture/float_types.rs: New.
* rust/compile/torture/forward_decl_1.rs: New.
* rust/compile/torture/forward_decl_2.rs: New.
* rust/compile/torture/forward_decl_3-unsafe.rs: New.
* rust/compile/torture/forward_decl_3.rs: New.
* rust/compile/torture/forward_decl_4.rs: New.
* rust/compile/torture/forward_decl_5.rs: New.
* rust/compile/torture/func1.rs: New.
* rust/compile/torture/func2.rs: New.
* rust/compile/torture/function_reference1.rs: New.
* rust/compile/torture/function_reference2.rs: New.
* rust/compile/torture/function_reference3.rs: New.
* rust/compile/torture/function_reference4.rs: New.
* rust/compile/torture/generics1.rs: New.
* rust/compile/torture/generics10.rs: New.
* rust/compile/torture/generics11.rs: New.
* rust/compile/torture/generics12.rs: New.
* rust/compile/torture/generics13.rs: New.
* rust/compile/torture/generics14.rs: New.
* rust/compile/torture/generics15.rs: New.
* rust/compile/torture/generics16.rs: New.
* rust/compile/torture/generics17.rs: New.
* rust/compile/torture/generics18.rs: New.
* rust/compile/torture/generics19.rs: New.
* rust/compile/torture/generics2.rs: New.
* rust/compile/torture/generics20.rs: New.
* rust/compile/torture/generics21.rs: New.
* rust/compile/torture/generics22.rs: New.
* rust/compile/torture/generics23.rs: New.
* rust/compile/torture/generics24.rs: New.
* rust/compile/torture/generics25.rs: New.
* rust/compile/torture/generics26.rs: New.
* rust/compile/torture/generics27.rs: New.
* rust/compile/torture/generics28.rs: New.
* rust/compile/torture/generics29.rs: New.
* rust/compile/torture/generics3.rs: New.
* rust/compile/torture/generics30.rs: New.
* rust/compile/torture/generics31.rs: New.
* rust/compile/torture/generics32.rs: New.
* rust/compile/torture/generics4.rs: New.
* rust/compile/torture/generics5.rs: New.
* rust/compile/torture/generics6.rs: New.
* rust/compile/torture/generics7.rs: New.
* rust/compile/torture/generics8.rs: New.
* rust/compile/torture/generics9.rs: New.
* rust/compile/torture/grouped_expr_function.rs: New.
* rust/compile/torture/identifier-missing-impl-1.rs: New.
* rust/compile/torture/if.rs: New.
* rust/compile/torture/if_elif.rs: New.
* rust/compile/torture/if_elif_else_expr1.rs: New.
* rust/compile/torture/if_else.rs: New.
* rust/compile/torture/ifunaryexpr.rs: New.
* rust/compile/torture/impl_block1.rs: New.
* rust/compile/torture/impl_block2.rs: New.
* rust/compile/torture/impl_block3.rs: New.
* rust/compile/torture/impl_block_unused.rs: New.
* rust/compile/torture/implicit_returns1.rs: New.
* rust/compile/torture/infer_type1.rs: New.
* rust/compile/torture/inner_attributes.rs: New.
* rust/compile/torture/integer_inference_var1.rs: New.
* rust/compile/torture/integer_inference_var2.rs: New.
* rust/compile/torture/integer_inference_var3.rs: New.
* rust/compile/torture/integer_inference_var4.rs: New.
* rust/compile/torture/integer_inference_var5.rs: New.
* rust/compile/torture/integer_types.rs: New.
* rust/compile/torture/intrinsics-1.rs: New.
* rust/compile/torture/intrinsics-2.rs: New.
* rust/compile/torture/isolated_cr_block_comment.rs: New.
* rust/compile/torture/isolated_cr_line_comment.rs: New.
* rust/compile/torture/issue-1024.rs: New.
* rust/compile/torture/issue-1075.rs: New.
* rust/compile/torture/issue-1432.rs: New.
* rust/compile/torture/issue-1434.rs: New.
* rust/compile/torture/issue-368.rs: New.
* rust/compile/torture/issue-808.rs: New.
* rust/compile/torture/issue-862.rs: New.
* rust/compile/torture/issue-893-2.rs: New.
* rust/compile/torture/issue-893.rs: New.
* rust/compile/torture/lazybooleanexpr_function.rs: New.
* rust/compile/torture/lifetime1.rs: New.
* rust/compile/torture/literals1.rs: New.
* rust/compile/torture/loop1.rs: New.
* rust/compile/torture/loop2.rs: New.
* rust/compile/torture/loop3.rs: New.
* rust/compile/torture/loop4.rs: New.
* rust/compile/torture/loop5.rs: New.
* rust/compile/torture/loop6.rs: New.
* rust/compile/torture/loop7.rs: New.
* rust/compile/torture/macro-issue1403.rs: New.
* rust/compile/torture/macro-issue1426.rs: New.
* rust/compile/torture/macro_as_expr.rs: New.
* rust/compile/torture/match1.rs: New.
* rust/compile/torture/methods1.rs: New.
* rust/compile/torture/methods2.rs: New.
* rust/compile/torture/methods3.rs: New.
* rust/compile/torture/mod-nameresolve.rs: New.
* rust/compile/torture/mod1.rs: New.
* rust/compile/torture/mod2.rs: New.
* rust/compile/torture/mod3.rs: New.
* rust/compile/torture/modules/mod.rs: New.
* rust/compile/torture/modules/valid_path.rs: New.
* rust/compile/torture/must_use1.rs: New.
* rust/compile/torture/must_use2.rs: New.
* rust/compile/torture/name_resolve1.rs: New.
* rust/compile/torture/negation_function.rs: New.
* rust/compile/torture/nested_fn1.rs: New.
* rust/compile/torture/nested_fn2.rs: New.
* rust/compile/torture/nested_struct1.rs: New.
* rust/compile/torture/never_type1.rs: New.
* rust/compile/torture/not_shebang.rs: New.
* rust/compile/torture/not_shebang_block_comment.rs: New.
* rust/compile/torture/not_shebang_comment.rs: New.
* rust/compile/torture/not_shebang_multiline_comment.rs: New.
* rust/compile/torture/not_shebang_spaces.rs: New.
* rust/compile/torture/parameter_usage1.rs: New.
* rust/compile/torture/parens1.rs: New.
* rust/compile/torture/pointer1.rs: New.
* rust/compile/torture/primconsts.rs: New.
* rust/compile/torture/prims_struct_eq.rs: New.
* rust/compile/torture/range-lang-item1.rs: New.
* rust/compile/torture/raw_identifiers.rs: New.
* rust/compile/torture/raw_identifiers_keywords.rs: New.
* rust/compile/torture/recursive_fn1.rs: New.
* rust/compile/torture/return_function.rs: New.
* rust/compile/torture/scoping1.rs: New.
* rust/compile/torture/self_type1.rs: New.
* rust/compile/torture/shadow1.rs: New.
* rust/compile/torture/shadow2.rs: New.
* rust/compile/torture/shebang.rs: New.
* rust/compile/torture/shebang_plus_attr.rs: New.
* rust/compile/torture/shebang_plus_attr2.rs: New.
* rust/compile/torture/static_function.rs: New.
* rust/compile/torture/static_var1.rs: New.
* rust/compile/torture/stmt_with_block1.rs: New.
* rust/compile/torture/str1.rs: New.
* rust/compile/torture/struct_access1.rs: New.
* rust/compile/torture/struct_base_init_1.rs: New.
* rust/compile/torture/struct_decl.rs: New.
* rust/compile/torture/struct_init.rs: New.
* rust/compile/torture/struct_init_10.rs: New.
* rust/compile/torture/struct_init_11.rs: New.
* rust/compile/torture/struct_init_2.rs: New.
* rust/compile/torture/struct_init_3.rs: New.
* rust/compile/torture/struct_init_4.rs: New.
* rust/compile/torture/struct_init_5.rs: New.
* rust/compile/torture/struct_init_6.rs: New.
* rust/compile/torture/struct_init_7.rs: New.
* rust/compile/torture/struct_init_8.rs: New.
* rust/compile/torture/struct_init_9.rs: New.
* rust/compile/torture/top_attr.rs: New.
* rust/compile/torture/traits1.rs: New.
* rust/compile/torture/traits10.rs: New.
* rust/compile/torture/traits11.rs: New.
* rust/compile/torture/traits12.rs: New.
* rust/compile/torture/traits13.rs: New.
* rust/compile/torture/traits14.rs: New.
* rust/compile/torture/traits15.rs: New.
* rust/compile/torture/traits16.rs: New.
* rust/compile/torture/traits17.rs: New.
* rust/compile/torture/traits18.rs: New.
* rust/compile/torture/traits19.rs: New.
* rust/compile/torture/traits2.rs: New.
* rust/compile/torture/traits3.rs: New.
* rust/compile/torture/traits4.rs: New.
* rust/compile/torture/traits5.rs: New.
* rust/compile/torture/traits6.rs: New.
* rust/compile/torture/traits7.rs: New.
* rust/compile/torture/traits8.rs: New.
* rust/compile/torture/traits9.rs: New.
* rust/compile/torture/transmute-size-check-1.rs: New.
* rust/compile/torture/transmute1.rs: New.
* rust/compile/torture/tuple1.rs: New.
* rust/compile/torture/tuple2.rs: New.
* rust/compile/torture/tuple3.rs: New.
* rust/compile/torture/tuple_enum_variants.rs: New.
* rust/compile/torture/tuple_field_access.rs: New.
* rust/compile/torture/tuple_function.rs: New.
* rust/compile/torture/tuple_index.rs: New.
* rust/compile/torture/tuple_struct1.rs: New.
* rust/compile/torture/tuple_struct2.rs: New.
* rust/compile/torture/tuple_struct_unit.rs: New.
* rust/compile/torture/tuple_struct_unused.rs: New.
* rust/compile/torture/type-alias1.rs: New.
* rust/compile/torture/type-alias2.rs: New.
* rust/compile/torture/type_infer1.rs: New.
* rust/compile/torture/type_infer2.rs: New.
* rust/compile/torture/type_infer3.rs: New.
* rust/compile/torture/type_infer4.rs: New.
* rust/compile/torture/type_infer5.rs: New.
* rust/compile/torture/type_infer6.rs: New.
* rust/compile/torture/unary_operators.rs: New.
* rust/compile/torture/undended-string-1.rs: New.
* rust/compile/torture/undended-string-2.rs: New.
* rust/compile/torture/underscore_id.rs: New.
* rust/compile/torture/union.rs: New.
* rust/compile/torture/union_union.rs: New.
* rust/compile/torture/unit_type1.rs: New.
* rust/compile/torture/unit_type2.rs: New.
* rust/compile/torture/unit_type3.rs: New.
* rust/compile/torture/unit_type4.rs: New.
* rust/compile/torture/unit_type5.rs: New.
* rust/compile/torture/unsafe1.rs: New.
* rust/compile/torture/unsafe2.rs: New.
* rust/compile/torture/unsafe3.rs: New.
* rust/compile/torture/unsafe4.rs: New.
* rust/compile/torture/unused.rs: New.
* rust/compile/torture/unused1.rs: New.
* rust/compile/torture/unused_struct.rs: New.
* rust/compile/torture/unused_struct_field.rs: New.
* rust/compile/torture/usize1.rs: New.
* rust/compile/torture/very-broken-attr-string.rs: New.
* rust/compile/torture/while_function.rs: New.
* rust/compile/traits1.rs: New.
* rust/compile/traits10.rs: New.
* rust/compile/traits11.rs: New.
* rust/compile/traits12.rs: New.
* rust/compile/traits2.rs: New.
* rust/compile/traits3.rs: New.
* rust/compile/traits4.rs: New.
* rust/compile/traits5.rs: New.
* rust/compile/traits6.rs: New.
* rust/compile/traits7.rs: New.
* rust/compile/traits8.rs: New.
* rust/compile/traits9.rs: New.
* rust/compile/tuple1.rs: New.
* rust/compile/tuple_struct1.rs: New.
* rust/compile/tuple_struct2.rs: New.
* rust/compile/tuple_struct3.rs: New.
* rust/compile/type-alias1.rs: New.
* rust/compile/type-bindings1.rs: New.
* rust/compile/unary_negation.rs: New.
* rust/compile/unary_not.rs: New.
* rust/compile/unconstrained_type_param.rs: New.
* rust/compile/unicode_escape.rs: New.
* rust/compile/unsafe1.rs: New.
* rust/compile/unsafe10.rs: New.
* rust/compile/unsafe2.rs: New.
* rust/compile/unsafe3.rs: New.
* rust/compile/unsafe4.rs: New.
* rust/compile/unsafe5.rs: New.
* rust/compile/unsafe6.rs: New.
* rust/compile/unsafe7.rs: New.
* rust/compile/unsafe8.rs: New.
* rust/compile/unsafe9.rs: New.
* rust/compile/unterminated_c_comment.rs: New.
* rust/compile/use_1.rs: New.
* rust/compile/usize1.rs: New.
* rust/compile/xfail/lifetime_param.rs: New.
* rust/compile/xfail/struct_field_vis.rs: New.
* rust/compile/xfail/xfail.exp: New.

Co-authored-by: Arthur Cohen <arthur.cohen@embecosm.com>
Co-authored-by: Thomas Schwinge <thomas@codesourcery.com>
Co-authored-by: Mark Wielaard <mark@klomp.org>
Co-authored-by: Marc Poulhiès <dkm@kataplop.net>
19 months agogccrs: Add link cases testsuite
Philip Herron [Tue, 23 Aug 2022 15:00:51 +0000 (16:00 +0100)]
gccrs: Add link cases testsuite

This testsuite is heavily inspired from the LTO testsuite that uses a
pattern where each file is compiled to an object file and finally linked
together. Since Rust does not have headers/prototypes, we rely on the
ordering here so that all files numbered greater than zero get compiled to
object files first. This leaves the _0 file free to test the 'extern crate' and
'use' keywords to force testing of the compiler to read metadata from the
other 'crates'.

gcc/testsuite/
* rust/link/generic_function_0.rs: New.
* rust/link/generic_function_1.rs: New.
* rust/link/link.exp: New.
* rust/link/simple_function_0.rs: New.
* rust/link/simple_function_1.rs: New.
* rust/link/trait_import_0.rs: New.
* rust/link/trait_import_1.rs: New.

19 months agogccrs: Add Debug info testsuite
Philip Herron [Tue, 23 Aug 2022 14:59:12 +0000 (15:59 +0100)]
gccrs: Add Debug info testsuite

This testsuite is specifically about testcases which scan the asm debug
info for results.

gcc/testsuite/
* rust/debug/chartype.rs: New.
* rust/debug/custom_link_section.rs: New.
* rust/debug/debug.exp: New.
* rust/debug/i8u8.rs: New.
* rust/debug/lang.rs: New.
* rust/debug/no_mangle.rs: New.
* rust/debug/oldlang.rs: New.
* rust/debug/tuple.rs: New.
* rust/debug/win64-abi.rs: New.

Co-authored-by: Tom Tromey <tom@tromey.com>
Co-authored-by: Mark Wielaard <mark@klomp.org>
Co-authored-by: Marc Poulhiès <dkm@kataplop.net>
19 months agogccrs: Add necessary hooks for a Rust front-end testsuite
Philip Herron [Tue, 23 Aug 2022 14:49:47 +0000 (15:49 +0100)]
gccrs: Add necessary hooks for a Rust front-end testsuite

This copies over code from other front-end testsuites to enable testing
for the rust front-end specifically.

gcc/testsuite/
* lib/rust-dg.exp: New.
* lib/rust.exp: New.

Co-authored-by: Marc Poulhiès <dkm@kataplop.net>
Co-authored-by: Thomas Schwinge <thomas@codesourcery.com>
19 months agoUse DW_ATE_UTF for the Rust 'char' type
Tom Tromey [Tue, 23 Aug 2022 14:47:13 +0000 (15:47 +0100)]
Use DW_ATE_UTF for the Rust 'char' type

The Rust 'char' type should use the DWARF DW_ATE_UTF encoding.

gcc/
* dwarf2out.cc (is_rust): New.
(base_type_die): Use DW_ATE_UTF for the Rust 'char' type.
(gen_compile_unit_die): Handle "GNU Rust".

Co-authored-by: Mark Wielaard <mark@klomp.org>
Co-authored-by: Marc Poulhiès <dkm@kataplop.net>
19 months agotree-optimization/108076 - if-conversion and forced labels
Richard Biener [Mon, 12 Dec 2022 16:52:46 +0000 (17:52 +0100)]
tree-optimization/108076 - if-conversion and forced labels

When doing if-conversion we simply throw away labels without checking
whether they are possibly targets of non-local gotos or have their
address taken.  The following rectifies this and refuses to if-convert
such loops.

PR tree-optimization/108076
* tree-if-conv.cc (if_convertible_loop_p_1): Reject blocks
with non-local or forced labels that we later remove
labels from.

* gcc.dg/torture/pr108076.c: New testcase.

19 months agolibsanitizer: Fix up libbacktrace build after r13-4547 [PR108072]
Jakub Jelinek [Tue, 13 Dec 2022 09:30:36 +0000 (10:30 +0100)]
libsanitizer: Fix up libbacktrace build after r13-4547 [PR108072]

The r13-4547 commit added new non-static function to libbacktrace:
backtrace_uncompress_zstd but for the libsanitizer use we need to
rename it, so that it is in __asan_* namespace and doesn't clash
with other copies of libbacktrace.

2022-12-13  Jakub Jelinek  <jakub@redhat.com>

libsanitizer/
PR sanitizer/108072
* libbacktrace/backtrace-rename.h (backtrace_uncompress_zstd): Define.

19 months agors6000: enable cbranchcc4
Haochen Gui [Tue, 13 Dec 2022 08:45:10 +0000 (16:45 +0800)]
rs6000: enable cbranchcc4

This patch enables "have_cbranchcc4" on rs6000 by defining a
"cbranchcc4" expander. "have_cbrnachcc4" is a flag in ifcvt.cc to
indicate if branching by CC bits is valid or not. With this flag
enabled, some branches can be optimized to conditional moves.

2022-12-07  Haochen Gui <guihaoc@linux.ibm.com>

gcc/
* config/rs6000/rs6000.md (cbranchcc4): New expander.

gcc/testsuite
* gcc.target/powerpc/cbranchcc4-1.c: New.
* gcc.target/powerpc/cbranchcc4-2.c: New.

19 months agooptabs: make prepare_cmp_insn goto fail when cbranchcc4 checks unsatisfied
Haochen Gui [Thu, 8 Dec 2022 05:22:29 +0000 (13:22 +0800)]
optabs: make prepare_cmp_insn goto fail when cbranchcc4 checks unsatisfied

prepare_cmp_insn is a help function to generate comparison rtx.
It should not assume that cbranchcc4 exists and all sub-CC modes
are supported on a target.  When the check fails, it could go to
fail and return a NULL rtx as its callers check the return value
for CCmode.

The test case (gcc.target/powerpc/cbranchcc4-1.c) which covers
failure path will be committed with an rs6000 specific patch.

2022-12-05  Haochen Gui <guihaoc@linux.ibm.com>

gcc/
* optabs.cc (prepare_cmp_insn): Return a NULL rtx other than
assertion failure when targets don't have cbranch optab or
predicate check fails.

19 months agoDaily bump.
GCC Administrator [Tue, 13 Dec 2022 00:18:07 +0000 (00:18 +0000)]
Daily bump.

19 months agoRevert "c++: build initializer_list<string> in a loop [PR105838]"
Jason Merrill [Mon, 12 Dec 2022 22:41:39 +0000 (17:41 -0500)]
Revert "c++: build initializer_list<string> in a loop [PR105838]"

This patch broke a couple of different patterns; reverting while I work on a
fix.

PR c++/108701

This reverts commit bd0485f20f4794f9787237706a6308473a8e9415.

19 months agolibgo: bump major version
Ian Lance Taylor [Mon, 12 Dec 2022 20:46:40 +0000 (12:46 -0800)]
libgo: bump major version

PR go/108057

The current version is the same as for the previous GCC release,
but there have been minor changes like new type descriptors that
make it impossible to run Go programs built with the previous GCC
release with the current libgo.

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/456976

19 months agoFortran: improve checking of assumed-size array spec [PR102180]
Harald Anlauf [Sun, 11 Dec 2022 22:24:03 +0000 (23:24 +0100)]
Fortran: improve checking of assumed-size array spec [PR102180]

gcc/fortran/ChangeLog:

PR fortran/102180
* array.cc (match_array_element_spec): Add check for bad
assumed-implied-spec.
(gfc_match_array_spec): Reorder logic so that the first bad array
element spec may trigger an error.

gcc/testsuite/ChangeLog:

PR fortran/102180
* gfortran.dg/pr102180.f90: New test.

19 months agod: Fix undefined reference to nested lambda in template (PR108055)
Iain Buclaw [Sat, 10 Dec 2022 21:11:41 +0000 (22:11 +0100)]
d: Fix undefined reference to nested lambda in template (PR108055)

Sometimes, nested lambdas of templated functions get no code generation
due to them being marked as instantianted outside of all modules being
compiled in the current compilation unit.  This despite enclosing
template instances being marked as instantiated inside the current
compilation unit.  To fix, all enclosing templates are now checked in
`function_defined_in_root_p'.

Because of this change, `function_needs_inline_definition_p' has also
been fixed up to only check whether the regular function definition
itself is to be emitted in the current compilation unit.

PR d/108055

gcc/d/ChangeLog:

* decl.cc (function_defined_in_root_p): Check all enclosing template
instances for definition in a root module.
(function_needs_inline_definition_p): Replace call to
function_defined_in_root_p with test for outer module `isRoot'.

gcc/testsuite/ChangeLog:

* gdc.dg/torture/imports/pr108055conv.d: New.
* gdc.dg/torture/imports/pr108055spec.d: New.
* gdc.dg/torture/imports/pr108055write.d: New.
* gdc.dg/torture/pr108055.d: New test.

19 months agoAArch64: Enable TARGET_CONST_ANCHOR
Wilco Dijkstra [Mon, 12 Dec 2022 15:44:03 +0000 (15:44 +0000)]
AArch64: Enable TARGET_CONST_ANCHOR

Enable TARGET_CONST_ANCHOR to allow complex constants to be created via
immediate add/sub.  Use a 24-bit range as that enables a 3 or 4-instruction
immediate to be replaced by 2 add/sub instructions.  Fix the costing of
add/sub to support 24-bit and 12-bit shifted immediates.
The generated code for the testcase is now the same or better than LLVM.
It also results in a small codesize reduction on SPEC.

gcc/
* config/aarch64/aarch64.cc (aarch64_rtx_costs): Add correct costs
for 24-bit and 12-bit shifted immediate add/sub.
(TARGET_CONST_ANCHOR): Define.
* config/aarch64/predicates.md (aarch64_pluslong_immediate):
Fix range check.

gcc/testsuite/
* gcc.target/aarch64/movk_3.c: New test.

19 months agomiddle-end: simplify complex if expressions where comparisons are inverse of one...
Tamar Christina [Mon, 12 Dec 2022 15:21:39 +0000 (15:21 +0000)]
middle-end: simplify complex if expressions where comparisons are inverse of one another.

This optimizes the following sequence

  ((a < b) & c) | ((a >= b) & d)

into

  (a < b ? c : d) & 1

for scalar and on vector we can omit the & 1.

Also recognizes

  (-(a < b) & c) | (-(a >= b) & d)

into

  a < b ? c : d

This changes the code generation from

zoo2:
cmp     w0, w1
cset    w0, lt
cset    w1, ge
and     w0, w0, w2
and     w1, w1, w3
orr     w0, w0, w1
ret

into

cmp w0, w1
csel w0, w2, w3, lt
and w0, w0, 1
ret

and significantly reduces the number of selects we have to do in the vector
code.

gcc/ChangeLog:

* match.pd: Add new rule.

gcc/testsuite/ChangeLog:

* gcc.target/aarch64/if-compare_1.c: New test.
* gcc.target/aarch64/if-compare_2.c: New test.

19 months agoAArch64: Fix vector re-interpretation between partial SIMD modes
Tamar Christina [Mon, 12 Dec 2022 15:20:30 +0000 (15:20 +0000)]
AArch64: Fix vector re-interpretation between partial SIMD modes

While writing a patch series I started getting incorrect codegen out from
VEC_PERM on partial struct types.

It turns out that this was happening because the TARGET_CAN_CHANGE_MODE_CLASS
implementation has a slight bug in it.  The hook only checked for SIMD to
Partial but never Partial to SIMD.   This resulted in incorrect subregs to be
generated from the fallback code in VEC_PERM_EXPR expansions.

I have unfortunately not been able to trigger it using a standalone testcase as
the mid-end optimizes away the permute every time I try to describe a permute
that would result in the bug.

The patch now rejects any conversion of partial SIMD struct types, unless they
are both partial structures of the same number of registers or one is a SIMD
type who's size is less than 8 bytes.

gcc/ChangeLog:

* config/aarch64/aarch64.cc (aarch64_can_change_mode_class): Restrict
conversions between partial struct types properly.

19 months agoAArch64: Support new tbranch optab.
Tamar Christina [Mon, 12 Dec 2022 15:18:56 +0000 (15:18 +0000)]
AArch64: Support new tbranch optab.

This implements the new tbranch optab for AArch64.

we cannot emit one big RTL for the final instruction immediately.
The reason that all comparisons in the AArch64 backend expand to separate CC
compares, and separate testing of the operands is for ifcvt.

The separate CC compare is needed so ifcvt can produce csel, cset etc from the
compares.  Unlike say combine, ifcvt can not do recog on a parallel with a
clobber.  Should we emit the instruction directly then ifcvt will not be able
to say, make a csel, because we have no patterns which handle zero_extract and
compare. (unlike combine ifcvt cannot transform the extract into an AND).

While you could provide various patterns for this (and I did try) you end up
with broken patterns because you can't add the clobber to the CC register.  If
you do, ifcvt recog fails.

i.e.

int
f1 (int x)
{
  if (x & 1)
    return 1;
  return x;
}

We lose csel here.

Secondly the reason the compare with an explicit CC mode is needed is so that
ifcvt can transform the operation into a version that doesn't require the flags
to be set.  But it only does so if it know the explicit usage of the CC reg.

For instance

int
foo (int a, int b)
{
  return ((a & (1 << 25)) ? 5 : 4);
}

Doesn't require a comparison, the optimal form is:

foo(int, int):
        ubfx    x0, x0, 25, 1
        add     w0, w0, 4
        ret

and no compare is actually needed.  If you represent the instruction using an
ANDS instead of a zero_extract then you get close, but you end up with an ands
followed by an add, which is a slower operation.

gcc/ChangeLog:

* config/aarch64/aarch64.md (*tb<optab><mode>1): Rename to...
(*tb<optab><ALLI:mode><GPI:mode>1): ... this.
(tbranch_<code><mode>4): New.
* config/aarch64/iterators.md(ZEROM, zerom): New.

gcc/testsuite/ChangeLog:

* gcc.target/aarch64/tbz_1.c: New test.

19 months agomiddle-end: Add new tbranch optab to add support for bit-test-and-branch operations
Tamar Christina [Mon, 12 Dec 2022 15:16:50 +0000 (15:16 +0000)]
middle-end: Add new tbranch optab to add support for bit-test-and-branch operations

This adds a new test-and-branch optab that can be used to do a conditional test
of a bit and branch.   This is similar to the cbranch optab but instead can
test any arbitrary bit inside the register.

This patch recognizes boolean comparisons and single bit mask tests.

gcc/ChangeLog:

* dojump.cc (do_jump): Pass along value.
(do_jump_by_parts_greater_rtx): Likewise.
(do_jump_by_parts_zero_rtx): Likewise.
(do_jump_by_parts_equality_rtx): Likewise.
(do_compare_rtx_and_jump): Likewise.
(do_compare_and_jump): Likewise.
* dojump.h (do_compare_rtx_and_jump): New.
* optabs.cc (emit_cmp_and_jump_insn_1): Refactor to take optab to check.
(validate_test_and_branch): New.
(emit_cmp_and_jump_insns): Optiobally take a value, and when value is
supplied then check if it's suitable for tbranch.
* optabs.def (tbranch_eq$a4, tbranch_ne$a4): New.
* doc/md.texi (tbranch_@var{op}@var{mode}4): Document it.
* optabs.h (emit_cmp_and_jump_insns): New.
* tree.h (tree_zero_one_valued_p): New.

19 months agoaarch64: Make existing V2HF be usable.
Tamar Christina [Mon, 12 Dec 2022 15:15:07 +0000 (15:15 +0000)]
aarch64: Make existing V2HF be usable.

The backend has an existing V2HFmode that is used by pairwise operations.
This mode was however never made fully functional.  Amongst other things it was
never declared as a vector type which made it unusable from the mid-end.

It's also lacking an implementation for load/stores so reload ICEs if this mode
is every used.  This finishes the implementation by providing the above.

Note that I have created a new iterator VHSDF_P instead of extending VHSDF
because the previous iterator is used in far more things than just load/stores.

It's also used for instance in intrinsics and extending this would force me to
provide support for mangling the type while we never expose it through
intrinsics.

gcc/ChangeLog:

* config/aarch64/aarch64-simd.md (*aarch64_simd_movv2hf): New.
(mov<mode>, movmisalign<mode>, aarch64_dup_lane<mode>,
aarch64_store_lane0<mode>, aarch64_simd_vec_set<mode>,
@aarch64_simd_vec_copy_lane<mode>, vec_set<mode>,
reduc_<optab>_scal_<mode>, reduc_<fmaxmin>_scal_<mode>,
aarch64_reduc_<optab>_internal<mode>, aarch64_get_lane<mode>,
vec_init<mode><Vel>, vec_extract<mode><Vel>): Support V2HF.
(aarch64_simd_dupv2hf): New.
* config/aarch64/aarch64.cc (aarch64_classify_vector_mode):
Add E_V2HFmode.
* config/aarch64/iterators.md (VHSDF_P): New.
(V2F, VMOVE, nunits, Vtype, Vmtype, Vetype, stype, VEL,
Vel, q, vp): Add V2HF.
* config/arm/types.md (neon_fp_reduc_add_h): New.

gcc/testsuite/ChangeLog:

* gcc.target/aarch64/sve/slp_1.c: Update testcase.

19 months agolibstdc++: Add a test checking for chrono::duration overflows
Jonathan Wakely [Mon, 12 Dec 2022 12:51:49 +0000 (12:51 +0000)]
libstdc++: Add a test checking for chrono::duration overflows

This test fails if chrono::days::rep or chrono::years::rep is a 32-bit
type, because a large days or years value silently overflows a 32-bit
integer when converted to seconds. It would be conforming to implement
chrono::days as chrono::duration<int32_t, ratio<86400>>, but would make
this overflow case more likely. Similarly for chrono::years,
chrono::months and chrono::weeks. This test is here to remind us not to
make that change lightly.

libstdc++-v3/ChangeLog:

* testsuite/20_util/duration/arithmetic/overflow_c++20.cc: New
test.

19 months agolibstdc++: Fix constraint on std::basic_format_string [PR108024]
Jonathan Wakely [Mon, 12 Dec 2022 11:40:07 +0000 (11:40 +0000)]
libstdc++: Fix constraint on std::basic_format_string [PR108024]

Also remove some redundant std::move calls for return statements.

libstdc++-v3/ChangeLog:

PR libstdc++/108024
* include/std/format (basic_format_string): Fix constraint.
* testsuite/std/format/format_string.cc: New test.

19 months agolibstdc++: Change names that clash with Win32 or Clang
Jonathan Wakely [Mon, 12 Dec 2022 11:22:45 +0000 (11:22 +0000)]
libstdc++: Change names that clash with Win32 or Clang

Clang now defines an __is_unsigned built-in, and Windows defines an
_Out_ macro. Replace uses of those as identifiers.

There might also be a problem with __is_signed, which we use in several
places.

libstdc++-v3/ChangeLog:

* include/std/chrono (hh_mm_ss): Rename __is_unsigned member to
_S_is_unsigned.
* include/std/format (basic_format_context): Rename _Out_
template parameter to _Out2.
* testsuite/17_intro/names.cc: Add Windows SAL annotation
macros.

19 months agolibstdc++: Define atomic lock-free type aliases for C++20 [PR98034]
Jonathan Wakely [Mon, 12 Dec 2022 12:49:40 +0000 (12:49 +0000)]
libstdc++: Define atomic lock-free type aliases for C++20 [PR98034]

libstdc++-v3/ChangeLog:

PR libstdc++/98034
* include/std/atomic (__cpp_lib_atomic_lock_free_type_aliases):
Define macro.
(atomic_signed_lock_free, atomic_unsigned_lock_free): Define
aliases.
* include/std/version (__cpp_lib_atomic_lock_free_type_aliases):
Define macro.
* testsuite/29_atomics/atomic/lock_free_aliases.cc: New test.

19 months agolibstdc++: Make operator<< for stacktraces less templated (LWG 3515)
Jonathan Wakely [Fri, 9 Dec 2022 14:59:01 +0000 (14:59 +0000)]
libstdc++: Make operator<< for stacktraces less templated (LWG 3515)

This change was approved for C++23 last month.

libstdc++-v3/ChangeLog:

* include/std/stacktrace (operator<<): Only output to narrow
ostreams (LWG 3515).
* testsuite/19_diagnostics/stacktrace/synopsis.cc:

19 months agomklog: do not parse binary file for PR entry
Martin Liska [Mon, 12 Dec 2022 12:24:25 +0000 (13:24 +0100)]
mklog: do not parse binary file for PR entry

contrib/ChangeLog:

* mklog.py: Do not search PR entry in a file that is binary.

19 months agoaarch64: Add __ARM_FEATURE_PAUTH and __ARM_FEATURE_BTI ACLE defines
Kyrylo Tkachov [Mon, 12 Dec 2022 11:07:45 +0000 (11:07 +0000)]
aarch64: Add __ARM_FEATURE_PAUTH and __ARM_FEATURE_BTI ACLE defines

Recent ACLE additions specified the __ARM_FEATURE_PAUTH and __ARM_FEATURE_BTI macros [1] that the compiler
should define when the pointer authentication and BTI instructions are available (and don't act as NOPs).
We've received requests to enable them in GCC for aarch64, similar to clang [2].
It's a fairly simple patch and should be non-intrusive at this stage.
Pointer authentication has its own "pauth" feature flag, whereas BTI depends on an architecture level
of Armv8.5-a or later.

Bootstrapped and tested on aarch64-none-linux-gnu.

[1] https://github.com/ARM-software/acle/blob/main/main/acle.md#pointer-authentication
[2] https://reviews.llvm.org/rG7d40baa82b1f272f68de63f3c4f68d970bdcd6ed

gcc/ChangeLog:

* config/aarch64/aarch64-c.cc (aarch64_update_cpp_builtins): Define
__ARM_FEATURE_PAUTH and __ARM_FEATURE_BTI when appropriate.
* config/aarch64/aarch64.h (TARGET_BTI): Define.

gcc/testsuite/ChangeLog:

* gcc.target/aarch64/acle/bti_def.c: New test.
* gcc.target/aarch64/acle/pauth_def.c: New test.

19 months agoRevert parts of ADDR_EXPR/CONSTRUCTOR treatment change in match.pd
Richard Biener [Mon, 12 Dec 2022 07:56:41 +0000 (08:56 +0100)]
Revert parts of ADDR_EXPR/CONSTRUCTOR treatment change in match.pd

This reverts the part that substitutes from the definition of an
SSA name to the capture, thus ADDR_EXPR@0 eventually yielding
&y_1->a[i_2] instead of _3.  That's because I didn't think of
how to deal with substituting @0 in the result pattern.  So
the following re-instantiates the SSA def CONSTRUCTOR handling
and in the ADDR_EXPR helpers used by match.pd handles SSA names
defined to ADDR_EXPRs transparently.

* genmatch.cc (dt_simplify::gen): Revert last change.
* match.pd: Revert simplification of CONSTUCTOR leaf handling.
(&x cmp SSA_NAME): Handle ADDR_EXPR in SSA defs.
* fold-const.cc (split_address_to_core_and_offset): Handle
ADDR_EXPRs in SSA defs.
(address_compare): Likewise.

19 months agotree-optimization/89317 - another pattern for &p->x != p + 4
Richard Biener [Mon, 12 Dec 2022 07:13:33 +0000 (08:13 +0100)]
tree-optimization/89317 - another pattern for &p->x != p + 4

As seen in the original testcase for PR89317 we are missing
comparison simplification patterns for &p->x != p + 4.  Fixed
by making an existing one apply.  To make the pattern apply
during CCP we need to simplify ccp_fold to not use GENERIC
folding of conditions but also use GIMPLE folding.

PR tree-optimization/89317
* tree-ssa-ccp.cc (ccp_fold): Handle GIMPLE_COND via
gimple_fold_stmt_to_constant_1.
* match.pd (&a != &a + c): Apply to pointer_plus with non-ADDR_EXPR
base as well.

* gcc.dg/tree-ssa/pr89317.c: Amend.

19 months agoDaily bump.
GCC Administrator [Mon, 12 Dec 2022 00:22:21 +0000 (00:22 +0000)]
Daily bump.

19 months agoFortran: fix ICE on bad use of statement function [PR107995]
Steve Kargl [Sat, 10 Dec 2022 21:17:15 +0000 (22:17 +0100)]
Fortran: fix ICE on bad use of statement function [PR107995]

gcc/fortran/ChangeLog:

PR fortran/107995
* interface.cc (gfc_check_dummy_characteristics): Reject statement
function dummy arguments.

gcc/testsuite/ChangeLog:

PR fortran/107995
* gfortran.dg/pr107995.f90: New test.

19 months agod: Fix internal compiler error: in visit, at d/imports.cc:72 (PR108050)
Iain Buclaw [Sat, 10 Dec 2022 18:12:43 +0000 (19:12 +0100)]
d: Fix internal compiler error: in visit, at d/imports.cc:72 (PR108050)

The visitor for lowering IMPORTED_DECLs did not have an override for
dealing with importing OverloadSet symbols.  This has now been
implemented in the code generator.

PR d/108050

gcc/d/ChangeLog:

* decl.cc (DeclVisitor::visit (Import *)): Handle build_import_decl
returning a TREE_LIST.
* imports.cc (ImportVisitor::visit (OverloadSet *)): New override.

gcc/testsuite/ChangeLog:

* gdc.dg/imports/pr108050/mod1.d: New.
* gdc.dg/imports/pr108050/mod2.d: New.
* gdc.dg/imports/pr108050/package.d: New.
* gdc.dg/pr108050.d: New test.

19 months agounidiff: use newline='\n' argument
Martin Liska [Sun, 11 Dec 2022 16:18:12 +0000 (17:18 +0100)]
unidiff: use newline='\n' argument

In order to support CR on a line, we need to open files
with newline='\n' as our line endings supposed to be of UNIX style.

contrib/ChangeLog:

* check_GNU_style.py: Use newline=\n.
* check_GNU_style_lib.py: Simplify.
* gcc-changelog/git_commit.py: Fix issues seen
Rust patchset.
* gcc-changelog/git_email.py: Use newline argument.
* gcc-changelog/test_email.py: New test.
* gcc-changelog/test_patches.txt: New test.
* mklog.py: Use newline argument.

19 months agod: Merge upstream dmd, druntime c8ae4adb2e, phobos 792c8b7c1.
Iain Buclaw [Fri, 9 Dec 2022 17:59:38 +0000 (18:59 +0100)]
d: Merge upstream dmd, druntime c8ae4adb2e, phobos 792c8b7c1.

D front-end changes:

- Import dmd v2.101.0.
- Deprecate the ability to call `__traits(getAttributes)' on
  overload sets.
- Deprecate non-empty `for' statement increment clause with no
  effect.
- Array literals assigned to `scope' array variables can now be
  allocated on the stack.

D runtime changes:

- Import druntime v2.101.0.

Phobos changes:

- Import phobos v2.101.0.

gcc/d/ChangeLog:

* dmd/MERGE: Merge upstream dmd c8ae4adb2e.
* typeinfo.cc (check_typeinfo_type): Update for new front-end
interface.
(TypeInfoVisitor::visit (TypeInfoStructDeclaration *)): Remove warning
that toHash() must be declared 'nothrow @safe`.

libphobos/ChangeLog:

* libdruntime/MERGE: Merge upstream druntime c8ae4adb2e.
* src/MERGE: Merge upstream phobos 792c8b7c1.

19 months agod: Expand bsr intrinsic as `clz(arg) ^ (argsize - 1)'
Iain Buclaw [Sat, 10 Dec 2022 16:17:35 +0000 (17:17 +0100)]
d: Expand bsr intrinsic as `clz(arg) ^ (argsize - 1)'

As well as removing unnecessary casts, this results in less temporaries
being generated during the initial gimple lowering pass.  Otherwise the
code generated is identical to the former intrinsic expansion.

gcc/d/ChangeLog:

* intrinsics.cc (expand_intrinsic_bsf): Fix comment.
(expand_intrinsic_bsr): Use BIT_XOR_EXPR instead of MINUS_EXPR.

19 months agotree-optimization/89317 - missed folding of (p + 4) - &p->d
Richard Biener [Sun, 11 Dec 2022 13:07:34 +0000 (14:07 +0100)]
tree-optimization/89317 - missed folding of (p + 4) - &p->d

The PR notices we fail to simplify

  a_4 = &x_3(D)->data;
  b_5 = x_3(D) + 16;
  _1 = b_5 - a_4;

together with the enabler handling ADDR_EXPR leafs in separate
stmts in match.pd the suggested patterns work.

PR tree-optimization/89317
* match.pd ((p + b) - &p->c -> b - offsetof(c)): New patterns.

* gcc.dg/tree-ssa/pr89317.c: New testcase.

19 months agoTreat ADDR_EXPR and CONSTRUCTOR as GIMPLE/GENERIC magically
Richard Biener [Sun, 11 Dec 2022 11:32:49 +0000 (12:32 +0100)]
Treat ADDR_EXPR and CONSTRUCTOR as GIMPLE/GENERIC magically

The following allows to match ADDR_EXPR for both the invariant
&a.b case as well as the &p->d case in a separate definition
transparently.  This also allows to remove the hack we employ
for CONSTRUCTOR which we handle for example with

 (match vec_same_elem_p
  CONSTRUCTOR@0
  (if (TREE_CODE (@0) == SSA_NAME
       && uniform_vector_p (gimple_assign_rhs1 (SSA_NAME_DEF_STMT (@0))))))

Note CONSTUCTORs always appear as separate definition in GIMPLE,
but I continue to play safe and ADDR_EXPRs are now matched in
both places where previously ADDR_EXPR@0 would have missed
the &p->x case.

This is a prerequesite for the PR89317 fix.

* genmatch.cc (dt_node::gen_kids): Handle ADDR_EXPR in both
the GENERIC and GIMPLE op position.
(dt_simplify::gen): Capture both GENERIC and GIMPLE op
position for ADDR_EXPR and CONSTRUCTOR.
* match.pd: Simplify CONSTRUCTOR leaf handling.

* gcc.dg/tree-ssa/forwprop-3.c: Adjust.
* g++.dg/tree-ssa/pr31146-2.C: Likewise.

19 months agotree-optimization/106904 - bogus -Wstringopt-overflow with vectors
Richard Biener [Wed, 7 Dec 2022 13:42:24 +0000 (14:42 +0100)]
tree-optimization/106904 - bogus -Wstringopt-overflow with vectors

The following avoids CSE of &ps->wp to &ps->wp.hwnd confusing
-Wstringopt-overflow by making sure to produce addresses to the
biggest container from vectorization.  For this I introduce
strip_zero_offset_components which turns &ps->wp.hwnd into
&(*ps) and use that to base the vector data references on.
That will also work for addresses with variable components,
alternatively emitting pointer arithmetic via calling
get_inner_reference and gimplifying that would be possible
but likely more intrusive.

This is by no means a complete fix for all of those issues
(avoiding ADDR_EXPRs in favor of pointer arithmetic might be).
Other passes will have similar issues.

In theory that might now cause false negatives.

PR tree-optimization/106904
* tree.h (strip_zero_offset_components): Declare.
* tree.cc (strip_zero_offset_components): Define.
* tree-vect-data-refs.cc (vect_create_addr_base_for_vector_ref):
Strip zero offset components before building the address.

* gcc.dg/Wstringop-overflow-pr106904.c: New testcase.

19 months agofortran/openmp.cc: Remove 's' that slipped in during %<..%> replacement
Tobias Burnus [Sun, 11 Dec 2022 10:47:55 +0000 (11:47 +0100)]
fortran/openmp.cc: Remove 's' that slipped in during %<..%> replacement

Seemingly, 's' (in VI that's the 's'ubstitute command) appeared verbatim in
a gfc_error message when to doing the '...' to %<...%> replacements in commit
r13-4590-g84f6f8a2a97f88be01e223c9c9dbab801a4f501f

gcc/fortran/
* openmp.cc (gfc_match_omp_context_selector_specification):
Remove spurious 's' in an error message.

19 months agoDaily bump.
GCC Administrator [Sun, 11 Dec 2022 00:17:43 +0000 (00:17 +0000)]
Daily bump.

19 months agoFortran: reject bad SIZE argument while simplifying ISHFTC [PR106911]
Harald Anlauf [Sat, 10 Dec 2022 21:05:15 +0000 (22:05 +0100)]
Fortran: reject bad SIZE argument while simplifying ISHFTC [PR106911]

gcc/fortran/ChangeLog:

PR fortran/106911
* simplify.cc (gfc_simplify_ishftc): If the SIZE argument is known
to be outside the allowed range, terminate simplification.

gcc/testsuite/ChangeLog:

PR fortran/106911
* gfortran.dg/pr106911.f90: New test.

19 months agoivopts: Fix IP_END handling for asm goto [PR107997]
Jakub Jelinek [Sat, 10 Dec 2022 15:50:39 +0000 (16:50 +0100)]
ivopts: Fix IP_END handling for asm goto [PR107997]

The following testcase ICEs, because the latch bb ends with
asm goto which has both fallthrough to the header and one or more labels
in the header too.  In that case there is just a single edge out of the
latch block, but still the asm goto is stmt_ends_bb_p statement, yet
ivopts decides to emit an IV bump at the IP_END position and inserts
it into the same bb as the asm goto after it, which then fails verification
(control flow in the middle of bb).

The following patch fixes it by splitting the latch -> header edge in that
case and inserting into the newly created bb, where split_edge ->
redirect_edge_and_branch is able to deal with this case correctly.

2022-12-10  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/107997
* tree-ssa-loop-ivopts.cc: Include cfganal.h.
(create_new_iv) <case IP_END>: If ip_end_pos bb is non-empty and ends
with a stmt which ends bb, instead of adding iv update after it split
the latch edge and insert iterator into the new latch bb.

* gcc.c-torture/compile/pr107997.c: New test.

19 months agolibgomp: Handle OpenMP's reverse offloads
Tobias Burnus [Sat, 10 Dec 2022 12:42:08 +0000 (13:42 +0100)]
libgomp: Handle OpenMP's reverse offloads

This commit enabled reverse offload for nvptx such that gomp_target_rev
actually gets called.  And it fills the latter function to do all of
the following: finding the host function to the device func ptr and
copying the arguments to the host, processing the mapping/firstprivate,
calling the host function, copying back the data and freeing as needed.

The data handling is made easier by assuming that all host variables
either existed before (and are in the mapping) or that those are
devices variables not yet available on the host. Thus, the reverse
mapping can do without refcounts etc. Note that the spec disallows
inside a target region device-affecting constructs other than target
plus ancestor device-modifier and it also limits the clauses permitted
on this construct.

For the function addresses, an additional splay tree is used; for
the lookup of mapped variables, the existing splay-tree is used.
Unfortunately, its data structure requires a full walk of the tree;
Additionally, the just mapped variables are recorded in a separate
data structure an extra lookup. While the lookup is slow, assuming
that only few variables get mapped in each reverse offload construct
and that reverse offload is the exception and not performance critical,
this seems to be acceptable.

libgomp/ChangeLog:

* libgomp.h (struct target_mem_desc): Predeclare; move
below after 'reverse_splay_tree_node' and add rev_array
member.
(struct reverse_splay_tree_key_s, reverse_splay_compare): New.
(reverse_splay_tree_node, reverse_splay_tree,
reverse_splay_tree_key): New typedef.
(struct gomp_device_descr): Add mem_map_rev member.
* oacc-host.c (host_dispatch): NULL init .mem_map_rev.
* plugin/plugin-nvptx.c (GOMP_OFFLOAD_get_num_devices): Claim
support for GOMP_REQUIRES_REVERSE_OFFLOAD.
* splay-tree.h (splay_tree_callback_stop): New typedef; like
splay_tree_callback but returning int not void.
(splay_tree_foreach_lazy): Define; like splay_tree_foreach but
taking splay_tree_callback_stop as argument.
* splay-tree.c (splay_tree_foreach_internal_lazy,
splay_tree_foreach_lazy): New; but early exit if callback returns
nonzero.
* target.c: Instatiate splay_tree_c with splay_tree_prefix 'reverse'.
(gomp_map_lookup_rev): New.
(gomp_load_image_to_device): Handle reverse-offload function
lookup table.
(gomp_unload_image_from_device): Free devicep->mem_map_rev.
(struct gomp_splay_tree_rev_lookup_data, gomp_splay_tree_rev_lookup,
gomp_map_rev_lookup, struct cpy_data, gomp_map_cdata_lookup_int,
gomp_map_cdata_lookup): New auxiliary structs and functions for
gomp_target_rev.
(gomp_target_rev): Implement reverse offloading and its mapping.
(gomp_target_init): Init current_device.mem_map_rev.root.
* testsuite/libgomp.fortran/reverse-offload-2.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-3.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-4.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-5.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-5a.f90: New test without
mapping of on-device allocated variables.

19 months agoAdd initial ChangeLogs for modula2.
Gaius Mulley [Sat, 10 Dec 2022 09:10:29 +0000 (09:10 +0000)]
Add initial ChangeLogs for modula2.

Add initial ChangeLog file in libgm2 and gcc/m2.

ChangeLog:

* libgm2: (New directory).
* libgm2/ChangeLog: (New file).

gcc/ChangeLog:

* m2: (New directory).
* m2/ChangeLog: (New file).

Signed-off-by: Gaius Mulley <gaiusmod2@gmail.com>
19 months agoAdd stub 'gcc/rust/ChangeLog'
Thomas Schwinge [Sat, 10 Dec 2022 07:33:22 +0000 (08:33 +0100)]
Add stub 'gcc/rust/ChangeLog'

19 months agoFortran: Replace simple '.' quotes by %<.%>
Tobias Burnus [Sat, 10 Dec 2022 07:34:04 +0000 (08:34 +0100)]
Fortran: Replace simple '.' quotes by %<.%>

When using %qs instead of '%s' or %<=%> instead of '=' looks nicer
by having nicer quotes and bold text, if the terminal supports it;
otherwise, plain quotes are used.

gcc/fortran/ChangeLog:

* match.cc (gfc_match_member_sep): Use %<...%> in gfc_error.
* openmp.cc (gfc_match_oacc_routine, gfc_match_omp_context_selector,
gfc_match_omp_context_selector_specification,
gfc_match_omp_declare_variant, resolve_omp_clauses): Likewise;
use %qs instead of '%s'.
* primary.cc (match_real_constant, gfc_match_varspec): Likewise.
* resolve.cc (gfc_resolve_formal_arglist, resolve_operator,
resolve_ordinary_assign): Likewise.

19 months agoPrepare 'contrib/gcc-changelog/git_commit.py' for GCC/Rust
Thomas Schwinge [Sat, 10 Dec 2022 06:27:55 +0000 (07:27 +0100)]
Prepare 'contrib/gcc-changelog/git_commit.py' for GCC/Rust

contrib/
* gcc-changelog/git_commit.py (default_changelog_locations): Add
'gcc/rust'.
(bug_components): Add 'rust'.

19 months agoAdd ChangeLog directories for modula2 into git_commit.py.
Gaius Mulley [Sat, 10 Dec 2022 02:21:53 +0000 (02:21 +0000)]
Add ChangeLog directories for modula2 into git_commit.py.

Prepare to add changelogs for the Modula2 front end by changing
the contrib git_commit.py script.

contrib/ChangeLog:

* gcc-changelog/git_commit.py (default_changelog_locations):
New entry for gcc/m2.  New entry for libgm2.

Signed-off-by: Gaius Mulley <gaiusmod2@gmail.com>
19 months agolibbacktrace: rewrite and simplify main zstd loop
Ian Lance Taylor [Sat, 10 Dec 2022 01:44:52 +0000 (17:44 -0800)]
libbacktrace: rewrite and simplify main zstd loop

* elf.c (ZSTD_TABLE_*): Use elf_zstd_fse_baseline_entry.
(ZSTD_ENCODE_BASELINE_BITS): Define.
(ZSTD_DECODE_BASELINE, ZSTD_DECODE_BASEBITS): Define.
(elf_zstd_literal_length_base): New static const array.
(elf_zstd_match_length_base): Likewise.
(struct elf_zstd_fse_baseline_entry): Define.
(elf_zstd_make_literal_baseline_fse): New static function.
(elf_zstd_make_offset_baseline_fse): Likewise.
(elf_zstd_make_match_baseline_fse): Likewise.
(print_table, main): Use elf_zstd_fse_baseline_entry.
(elf_zstd_lit_table, elf_zstd_match_table): Likewise.
(elf_zstd_offset_table): Likewise.
(struct elf_zstd_seq_decode): Likewise.  Remove use_rle and rle
fields.
(elf_zstd_unpack_seq_decode): Use elf_zstd_fse_baseline_entry,
taking a conversion function.  Convert RLE to FSE.
(elf_zstd_literal_length_baseline): Remove.
(elf_zstd_literal_length_bits): Remove.
(elf_zstd_match_length_baseline): Remove.
(elf_zstd_match_length_bits): Remove.
(elf_zstd_decompress): Use elf_zstd_fse_baseline_entry.  Rewrite
and simplify main loop.

19 months agoDaily bump.
GCC Administrator [Sat, 10 Dec 2022 00:17:39 +0000 (00:17 +0000)]
Daily bump.

19 months agoFortran: ICE on recursive derived types with allocatable components [PR107872]
Paul Thomas [Fri, 9 Dec 2022 21:13:45 +0000 (22:13 +0100)]
Fortran: ICE on recursive derived types with allocatable components [PR107872]

gcc/fortran/ChangeLog:

PR fortran/107872
* resolve.cc (derived_inaccessible): Skip over allocatable components
to prevent an infinite loop.

gcc/testsuite/ChangeLog:

PR fortran/107872
* gfortran.dg/pr107872.f90: New test.

19 months agoFortran/OpenMP: align/allocator modifiers to the allocate clause
Tobias Burnus [Fri, 9 Dec 2022 20:45:37 +0000 (21:45 +0100)]
Fortran/OpenMP: align/allocator modifiers to the allocate clause

gcc/fortran/ChangeLog:

* dump-parse-tree.cc (show_omp_namelist): Improve OMP_LIST_ALLOCATE
output.
* gfortran.h (struct gfc_omp_namelist): Add 'align' to 'u'.
(gfc_free_omp_namelist): Add bool arg.
* match.cc (gfc_free_omp_namelist): Likewise; free 'u.align'.
* openmp.cc (gfc_free_omp_clauses, gfc_match_omp_clause_reduction,
gfc_match_omp_flush): Update call.
(gfc_match_omp_clauses): Match 'align/allocate modifers in
'allocate' clause.
(resolve_omp_clauses): Resolve align.
* st.cc (gfc_free_statement): Update call
* trans-openmp.cc (gfc_trans_omp_clauses): Handle 'align'.

libgomp/ChangeLog:

* libgomp.texi (5.1 Impl. Status): Split allocate clause/directive
item about 'align'; mark clause as 'Y' and directive as 'N'.
* testsuite/libgomp.fortran/allocate-2.f90: New test.
* testsuite/libgomp.fortran/allocate-3.f90: New test.

19 months agors6000: Remove useless copy_rtx in rs6000_emit_set_{,long}_const
Jiufu Guo [Fri, 9 Dec 2022 05:50:37 +0000 (13:50 +0800)]
rs6000: Remove useless copy_rtx in rs6000_emit_set_{,long}_const

Function rs6000_emit_set_const/rs6000_emit_set_long_const are only invoked from
two "define_split"s where the target operand is limited to gpc_reg_operand or
int_reg_operand, then the operand must be REG_P.
And in rs6000_emit_set_const/rs6000_emit_set_long_const, to create temp rtx,
it is using code like "gen_reg_rtx({S|D}Imode)", it must also be REG_P.
So, copy_rtx is not needed for temp and dest.

This patch removes those "copy_rtx" for rs6000_emit_set_const and
rs6000_emit_set_long_const.

gcc/ChangeLog:

* config/rs6000/rs6000.cc (rs6000_emit_set_const): Remove copy_rtx.
(rs6000_emit_set_long_const): Likewise.

19 months agoMAINTAINERS: fix spacing
Martin Liska [Fri, 9 Dec 2022 10:08:55 +0000 (11:08 +0100)]
MAINTAINERS: fix spacing

ChangeLog:

* MAINTAINERS: Fix spacing.

19 months agoi386: fix assert (__builtin_cpu_supports ("x86-64") >= 0)
Martin Liska [Fri, 25 Nov 2022 12:05:56 +0000 (13:05 +0100)]
i386: fix assert (__builtin_cpu_supports ("x86-64") >= 0)

Similar story as PR103661, we again return a negative number
for __builtin_cpu_supports:

Documentation says:

int __builtin_cpu_supports(const char *feature)
This function returns a positive integer if the run-time CPU supports feature and returns 0 otherwise.
while we return -2147483648.

Moreover, I noticed "x86-64" is not a valid option for __builtin_cpu_is,
but for __builtin_cpu_supports.

PR target/107551

gcc/ChangeLog:

* config/i386/i386-builtins.cc (fold_builtin_cpu): Use same path
as for PR103661.
* doc/extend.texi: Fix "x86-64" use.

gcc/testsuite/ChangeLog:

* gcc.target/i386/builtin_target.c: Add more checks.

19 months agoRename SUBTARGET_CC1_SPEC to OS_CC1_SPEC
Sebastian Huber [Fri, 9 Dec 2022 06:55:52 +0000 (07:55 +0100)]
Rename SUBTARGET_CC1_SPEC to OS_CC1_SPEC

This change resolves a naming conflict introduced by the recently added
SUBTARGET_CC1_SPEC to gcc.cc.  Some targets (mips and loongarch) aready used
a SUBTARGET_CC1_SPEC define.  Rename the define used by gcc.cc to OS_CC1_SPEC.

gcc/ChangeLog:

* config/rtems.h (SUBTARGET_CC1_SPEC): Rename to...
(OS_CC1_SPEC): ...this.
* gcc.cc (SUBTARGET_CC1_SPEC): Rename to...
(OS_CC1_SPEC): ...this.

19 months agoanalyzer: rename region-model-impl-calls.cc to kf.cc
David Malcolm [Fri, 9 Dec 2022 02:19:24 +0000 (21:19 -0500)]
analyzer: rename region-model-impl-calls.cc to kf.cc

gcc/ChangeLog:
* Makefile.in (ANALYZER_OBJS): Update for renaming of
analyzer/region-model-impl-calls.cc to analyzer/kf.cc.

gcc/analyzer/ChangeLog:
* analyzer.h (class known_function): Expand comment.
* region-model-impl-calls.cc: Rename to...
* kf.cc: ...this.
* known-function-manager.h (class known_function_manager): Add
leading comment.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>
19 months agoanalyzer: fix ICE on region creation during get_referenced_base_regions [PR108003]
David Malcolm [Fri, 9 Dec 2022 02:19:23 +0000 (21:19 -0500)]
analyzer: fix ICE on region creation during get_referenced_base_regions [PR108003]

gcc/analyzer/ChangeLog:
PR analyzer/108003
* call-summary.cc
(call_summary_replay::convert_region_from_summary_1): Convert
heap_regs_in_use from auto_sbitmap to auto_bitmap.
* region-model-manager.cc
(region_model_manager::get_or_create_region_for_heap_alloc):
Convert from sbitmap to bitmap.
* region-model-manager.h: Likewise.
* region-model.cc
(region_model::get_or_create_region_for_heap_alloc): Convert from
auto_sbitmap to auto_bitmap.
(region_model::get_referenced_base_regions): Likewise.
* region-model.h: Include "bitmap.h" rather than "sbitmap.h".
(region_model::get_referenced_base_regions): Convert from
auto_sbitmap to auto_bitmap.

gcc/testsuite/ChangeLog:
PR analyzer/108003
* g++.dg/analyzer/pr108003.C: New test.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>
19 months agoanalyzer: handle memmove like memcpy
David Malcolm [Fri, 9 Dec 2022 02:19:23 +0000 (21:19 -0500)]
analyzer: handle memmove like memcpy

gcc/analyzer/ChangeLog:
* region-model-impl-calls.cc (class kf_memcpy): Rename to...
(class kf_memcpy_memmove): ...this.
(kf_memcpy::impl_call_pre): Rename to...
(kf_memcpy_memmove::impl_call_pre): ...this, and check the src for
poison.
(register_known_functions): Update for above renaming, and
register BUILT_IN_MEMMOVE and BUILT_IN_MEMMOVE_CHK.

gcc/testsuite/ChangeLog:
* gcc.dg/analyzer/memcpy-1.c (test_8a, test_8b): New tests.
* gcc.dg/analyzer/memmove-1.c: New test, based on memcpy-1.c
* gcc.dg/analyzer/out-of-bounds-1.c (test7): Update expected
result for uninit srcBuf.
* gcc.dg/analyzer/out-of-bounds-5.c (test8, test9): Add
dg-warnings for memcpy from uninit src vla.
* gcc.dg/analyzer/pr104308.c (test_memmove_within_uninit):
Expect creation point note to be missing on riscv*-*-*.

Signed-off-by: David Malcolm <dmalcolm@redhat.com>
19 months agoEnable hwasan for x86-64.
liuhongt [Wed, 21 Apr 2021 21:23:48 +0000 (14:23 -0700)]
Enable hwasan for x86-64.

libsanitizer
* configure.tgt: Enable hwasan for x86-64.

19 months agoImplement hwasan target_hook.
liuhongt [Wed, 21 Apr 2021 21:15:04 +0000 (14:15 -0700)]
Implement hwasan target_hook.

gcc/ChangeLog:

* doc/invoke.texi (x86 options): Document
-mlam={none,u48,u57}.
* config/i386/i386-opts.h (enum lam_type): New enum.
* config/i386/i386.cc (ix86_memtag_can_tag_addresses): New.
(ix86_memtag_set_tag): Ditto.
(ix86_memtag_extract_tag): Ditto.
(ix86_memtag_add_tag): Ditto.
(ix86_memtag_tag_size): Ditto.
(ix86_memtag_untagged_pointer): Ditto.
(TARGET_MEMTAG_CAN_TAG_ADDRESSES): New.
(TARGET_MEMTAG_ADD_TAG): Ditto.
(TARGET_MEMTAG_SET_TAG): Ditto.
(TARGET_MEMTAG_EXTRACT_TAG): Ditto.
(TARGET_MEMTAG_UNTAGGED_POINTER): Ditto.
(TARGET_MEMTAG_TAG_SIZE): Ditto.
(IX86_HWASAN_SHIFT): Ditto.
(IX86_HWASAN_TAG_SIZE): Ditto.
* config/i386/i386-expand.cc (ix86_expand_call): Untag code
pointer.
* config/i386/i386-options.cc (ix86_option_override_internal):
Error when enable -mlam=[u48|u57] for 32-bit code.
* config/i386/i386.opt: Add -mlam=[none|u48|u57].
* config/i386/i386-protos.h (ix86_memtag_untagged_pointer):
Declare.
(ix86_memtag_can_tag_addresses): Ditto.

19 months agolibstdc++: Remove digit separators [PR108015]
Jonathan Wakely [Thu, 8 Dec 2022 11:55:31 +0000 (11:55 +0000)]
libstdc++: Remove digit separators [PR108015]

These are not valid in C++11 and cause a warning when preprocessing,
even though they're inside a skipped group.

chrono:2436: warning: missing terminating ' character

libstdc++-v3/ChangeLog:

PR libstdc++/108015
* include/std/chrono (hh_mm_ss): Remove digit separators.

19 months agolibstdc++: Fix some -Wunused warnings in tests
Jonathan Wakely [Thu, 8 Dec 2022 11:39:20 +0000 (11:39 +0000)]
libstdc++: Fix some -Wunused warnings in tests

libstdc++-v3/ChangeLog:

* include/ext/pb_ds/detail/type_utils.hpp (PB_DS_STATIC_ASSERT):
Add unused attribute to avoid -Wunused-local-typedef warnings.
* testsuite/17_intro/tag_type_explicit_ctor.cc: Add pragma to
ignore -Wunused-variable warnings

19 months agolibstdc++: Add [[nodiscard]] to chrono conversion functions
Jonathan Wakely [Mon, 28 Nov 2022 11:22:24 +0000 (11:22 +0000)]
libstdc++: Add [[nodiscard]] to chrono conversion functions

Also add doxygen comments.

libstdc++-v3/ChangeLog:

* include/bits/chrono.h (duration_cast, floor, round, abs, ceil)
(time_point_cast): Add [[nodiscard]] attribute and doxygen
comments.
(treat_as_floating_point): Add doxygen commen.

19 months agolibstdc++: Change class-key for duration and time_point to class
Jonathan Wakely [Fri, 2 Dec 2022 16:18:43 +0000 (16:18 +0000)]
libstdc++: Change class-key for duration and time_point to class

We define these with the 'struct' keyword, but the standard uses
'class'. This results in warnings if users try to refer to them using
elaborated type specifiers.

libstdc++-v3/ChangeLog:

* include/bits/chrono.h (duration, time_point): Change 'struct'
to 'class'.

19 months agoDaily bump.
GCC Administrator [Fri, 9 Dec 2022 00:18:54 +0000 (00:18 +0000)]
Daily bump.