If we fail to evaluate the kind expression we were getting an invalid
optional reference. Instead, fail with an assert. This can happen if an
intrinsic function is not folded, but that will be implemented
eventually.
Original-commit: flang-compiler/f18@3bdbfc34bb
Reviewed-on: https://github.com/flang-compiler/f18/pull/256
Tree-same-pre-rewrite: false
These are recognized along with other attributes and saved in
`passName_` and `bindName_`. The functions `SetPassNameOn()` and
`SetBindNameOn()` set them in a symbol if they are present.
They are also written to `.mod` files.
Add `MakePlaceholder()` to make symbols for names that otherwise
wouldn't have one. This allows us to assign a symbol to every name
and report errors for those that don't have one. Make use of this
for PASS names, which don't have explicit symbols.
Change `ObjectEntityDetails` and `ProcEntityDetails` to be sub-classes
of `EntityDetails`. They each contain a superset of the information in
`EntityDetails` so this reduces some duplication.
Original-commit: flang-compiler/f18@404c920840
Reviewed-on: https://github.com/flang-compiler/f18/pull/256
Tree-same-pre-rewrite: false
of ExpressionBase::derived(). gcc-8.2 on macOS was choosing to create
non-inline instances of derived() during the explicit instantiations of
ExpressionBase in expression.cc and fold.cc. During linking of any
executable, the linker failed when it found these duplicate definitions.
While this solution works, it removes the opportunity to inline the trivial
derived() functions. Another solution would be to make all of the
templates related to ExpressionBase in expression.cc and fold.cc available
in a single .cc file, where the explicit instantiation
FOR_EACH_TYPE_AND_KIND(template class ExpressionBase) is done once. This
approach would allow inlining, but require something like template
implementation headers that could be included into the instantiation .cc
file.
Original-commit: flang-compiler/f18@074de39418
Reviewed-on: https://github.com/flang-compiler/f18/pull/248
Tree-same-pre-rewrite: false
As with array constructors, we can encounter a decl-type-spec for a
structure-constructor while the decl-type-spec for the
type-declaration-stmt is still active. So here too we need to save and
restore the state.
In this case, it is done in DeclarationVisitor, so add
SetDeclTypeSpecState() to get and set the state.
Example where this occurs:
```
type(t(2)) :: x = t(2)()
```
Original-commit: flang-compiler/f18@aa5ca8d7b2
Reviewed-on: https://github.com/flang-compiler/f18/pull/249
Tree-same-pre-rewrite: false
When we encounter a ProcedureDesignator, if it is an external procedure
(i.e. there is no visible declaration) it must be implicitly created in
the global scope, not the current scope.
If there is already a symbol there (i.e. a module) it is an error, even
if that symbol is not visible due to an IMPORT statement.
Original-commit: flang-compiler/f18@835e234ded
Reviewed-on: https://github.com/flang-compiler/f18/pull/249
Tree-same-pre-rewrite: false
We can encounter a decl-type-spec for an array constructor while the
decl-type-spec for type-declaration-stmt is still active:
```
integer, allocatable :: a(:) = [integer :: 1, 2, 3]
```
To handle this, save the state of the current decl-type-spec while
processing the second one, then restore when done.
Original-commit: flang-compiler/f18@d125a0f612
Reviewed-on: https://github.com/flang-compiler/f18/pull/249
Tree-same-pre-rewrite: false
We need to save the names of specific procedures that appear in a
ProcedureStmt or GenericStmt to resolve at the end of the specification
section. They may be forward references to interface bodies.
We were saving them in the GenericDetails but that was not the right
place: we need to save the parser::Name so that we can eventually
resolve it to a Symbol, but the symbol table should not contains
references to Name because they go away with the parse tree.
The fix is to save the mapping in a new multimap, specificProcs_ in
InterfaceVisitor. We can reference parser::Names there and resolve them
during ResolveSpecificsInGenerics(). We no longer need to clutter
GenericDetails with data structures for unresolved names.
Also handle the case where a specific procedure is use-associated from
another module.
Original-commit: flang-compiler/f18@c7f7b1e72d
Reviewed-on: https://github.com/flang-compiler/f18/pull/249
Tree-same-pre-rewrite: false
Character decl-type-specs with no kind specified had no kind written to
the .mod file. That was different from the other intrinsic types which
always have an explicit type. This change make character like the other
intrinsic types.
Original-commit: flang-compiler/f18@5b4f047a13
Reviewed-on: https://github.com/flang-compiler/f18/pull/247
Tree-same-pre-rewrite: false
IntrinsicTypeSpec was used for all intrinsics except for character.
Change it to be a common base class for NumericTypeSpec,
LogicalTypeSpec, and CharacterTypeSpec.
Change DeclTypeSpec to categorize the intrinsics as Numeric, Logical,
and Character. Add some utility methods: AsIntrinsic() and IsNumeric().
In scope.h, give the functions that create DeclTypeSpecs better names.
In semantics.h, replace MakeIntrinsicTypeSpec() with MakeNumericType()
and MakeLogicalType() as it does not apply to character types.
Original-commit: flang-compiler/f18@8ad92d069c
Reviewed-on: https://github.com/flang-compiler/f18/pull/247
Tree-same-pre-rewrite: false
Recognize the various ways of specifying character lengths.
Define CharacterTypeSpec with length and kind and store them in the
current scope, as is done with DerivedTypeSpec (which can also have
length parameters).
Note: IntrinsicTypeSpec is no longer used for characters, so it
should have a different name. Similarly, in DeclTypeSpec::Category,
Intrinsic does not include Character.
Original-commit: flang-compiler/f18@5f84785193
Reviewed-on: https://github.com/flang-compiler/f18/pull/247
Tree-same-pre-rewrite: false
Now DeclTypeSpecs are stored in the scope so that they remain available
as long as the scope exists. DeclTypeSpecs for intrinsic types are
stored in the global scope; those for derived types are in the current
scope. They can contains type parameter values so they can't be reused.
Add `Semantics::MakeIntrinsicTypeSpec` to simplify creating
DeclTypeSpecs for intrinsic types.
Replace `std::optional<DeclTypeSpec>` with `DeclTypeSpec *` as they do
not need to be copied around.
Also fix a small bug with writing `class(t(...))` to the module file --
the type parameters were missing.
Original-commit: flang-compiler/f18@e4744418fc
Reviewed-on: https://github.com/flang-compiler/f18/pull/247
Tree-same-pre-rewrite: false
Type parameters were sorted by the order of the type-param-def-stmts.
But we need to preserve the order of the type-param-name-list.
The is the order of positional parameters in a derived-type-spec.
So add `paramNames` to `DerivedTypeDetails` to preserve the original
order. Using this allows us to write module files with both the
type-param-name-list and type-param-def-stmts in the original order.
Also fix a bug where a duplicate type-param-def caused a spurious
extra error. If `MakeTypeSymbol()` reports an error we should not
call `SetType()` because it will just report another error.
Original-commit: flang-compiler/f18@3ca55b6333
Reviewed-on: https://github.com/flang-compiler/f18/pull/239
This fixes a problem with converting the ubound call in the example
below back to Fortran (in this case, for writing to the .mod file).
One of the ActualArguments encountered in ProcedureRef::AsFortran is
a `std::nullopt`, so we need to handle that case.
```
module m
real :: x(10)
real :: y(ubound(x, dim=1))
end module
```
Original-commit: flang-compiler/f18@c5ace6b824
Reviewed-on: https://github.com/flang-compiler/f18/pull/240
It's not good enough to evaluate expressions in the symbol table after
name resolution has completed. This is because we need the values of
constant expressions for types, for example, we need to evaluate `k` in
`integer(k) :: x` to know the type of `x`.
So, eliminate `LazyExpr` and call `EvaluateExpr()` on expressions that
we need in the symbol table. The latter evaluates and folds an
expression in the current context. This is now possible because symbols
are added to `parser::Name` as soon as possible rather than in a pass
after name resolution. Along with `LazyExpr` we can eliminate the whole
`ResolveSymbolExprs` pass that used to resolve them.
In resolve-names.cc, many `Pre` functions are changed to `Post` so that
names are resolved before doing the associated processing. For example,
with intrinsic type specs, names in the kind expression must be resolved
before attempting to evaluate that expression.
In `GetSymbolType()` in type.cc, handle both `ObjectEntityDetails` and
`EntityDetails` by using `Symbol::GetType()`.
Add explicit declarations in label01.F90 because we can't handle
implicitly typed array bounds yet.
Original-commit: flang-compiler/f18@d67716640b
Reviewed-on: https://github.com/flang-compiler/f18/pull/238
We need access to the SemanticsContext and Walk() function everywhere,
so move them to a BaseVisitor, a base class at the root of the class
hierarchy.
Also move MessageHandler to be a data member of BaseVisitor and forward
Say calls to it.
Original-commit: flang-compiler/f18@eb9adc342a
Reviewed-on: https://github.com/flang-compiler/f18/pull/238
Tree-same-pre-rewrite: false
Rework how `parser::Name` is resolved to contain a `Symbol`. so that
constants in types can be evaluated. For example:
```
integer, parameter :: k = 8
integer(k) :: i
```
The old approach of collecting the symbols at the end of name resolution
and filling in the `parser::Name` does not work because the type of `i`
needs to be set in the symbol table.
The symbol field in `parser::Name` is now mutable so that we can set it
during name resolution. `RewriteParseTree` no longer needs to do that
(it still warns about unresolved ones), so it does not need to collect
symbols and fill them in. Consequently, we can eliminate "occurrences"
from symbols -- we just need the name where each is first defined.
This requires a lot of refactoring in `resolve-names.cc` to pass around
`parser::Name` rather than `SourceName` so that we can resolve the
name to a symbol.
Fix some bugs where we stored `SourceName *` instead of `SourceName`
in the symbol table. The pointers were into the parse tree, so they
were only valid as long as the parse tree was around. The symbol
table needs to remain valid longer than that, so the names need to
be copied. `parser::Name` is not used in the symbol table.
Eliminate `GenericSpec`. Currently all we need to do is to resolve
the kinds of GenericSpec that contain names.
Add `ScopeName` kind of `MiscDetails` for when we need a symbol in
the scope to match the name of the scope. For example, `module m`
cannot contain a declaration of a new `m`. Subprograms need real
details because they can be called recursively.
Fix output of partially resolved modules where we know it is a submodule
but have not yet resolved the ancestor.
Original-commit: flang-compiler/f18@5c1a4b99d2
Reviewed-on: https://github.com/flang-compiler/f18/pull/238
Tree-same-pre-rewrite: false
