template. The injected-class-name is either a type or a template,
depending on whether a '<' follows it. As a type, the
injected-class-name's template argument list contains its template
parameters in declaration order.
As part of this, add logic for canonicalizing declarations, and be
sure to canonicalize declarations used in template names and template
arguments.
A TagType is dependent if the declaration it references is dependent.
I'm not happy about the rather complicated protocol needed to use
ASTContext::getTemplateSpecializationType.
llvm-svn: 71408
compensating for super classes). This was making the reported class
sizes for empty classes very, very wrong.
- Also, we now report the size info for an empty class like gcc (as
the offset of the start, not as 0, 0).
- Add a few more test cases we were mishandling before (padding bit
field at end of struct, for example).
llvm-svn: 70938
- This implements gcc style Objective-C interface layout (I
think). Currently it is always off, there is no functionality
change unless this is passed.
For the curious, the deal is that gcc lays out the fields of a
subclass as if they were part of the superclass. That is, the
subclass fields immediately follow the super class fields instead
of being padded to the alignment of the superclass structure.
- Currently gcc uses the tight layout in 32-bit and 64-bit modes, and
llvm-gcc uses it in 32-bit only, for reasons which aren't clear
yet. We probably want to switch to matching gcc, once this makes it
through testing... my hope is that we can also fix llvm-gcc in
order to maintain compatibility between the compilers.
llvm-svn: 70827
via CollectObjCIvars.
- In places where we need them, we should have the implementation and
access the properties through it.
This is a fairly substantial functionality change:
1. @encode no longer encodes synthesized ivars, ever.
2. The ivar layout bitmap no longer encodes information for
synthesized ivars in superclasses. Well, actually I had already
broken that, but it is intentional now.
We are now differing substantially from llvm-gcc and gcc
here. However, in my opinion this fundamentally *must* work if
non-fragile classes are to work. Without this change, the result of
@encode and the ivar layout depend on the order that the
implementation is seen in a file (if it is in the same file with its
superclass). Since both scenarios should work the same, our behavior
is now consistent with gcc behavior as if an implementation is never
seen following an implementation of its superclass.
Note that #2 is only a functionality change when (A) an
implementation appears in the same translation unit with the
implementation of its superclass, and (B) the superclass has
synthesized ivars. My belief is that this situation does not occur in
practice.
I am not yet sure of the role/semantics of @encode when synthesized
ivars are present... it's use is fairly unsound in a non-fragile world.
llvm-svn: 70822
struct.
- We still need to do more lookup than necessary because ivars don't
live in a reasonable DeclContext.
- The only remaining client of the interface shadow struct is the
ivar layout bitmap.
llvm-svn: 70756
- These routines should now be independent of the Sema state.
- This is nearly zero functionality change, the distinction only
matters in the non-fragile ABI, and the consumers that care about
this distinction should be using getASTObjCImplementationLayout.
llvm-svn: 70692
- The difference from getASTObjCInterfaceLayout is that the computes
the layout including synthesized ivars.
- No functionality change, they currently both compute the same thing
-- whether that includes synthesized ivars or not depends on when
they get called!!!
llvm-svn: 70690
"aligned" attribute. Previously, we were skipping over these
attributes when we jumped directly to the canonical type. Now,
ASTContext::getTypeInfo walks through typedefs and other
"non-canonical" types manually, looking for "aligned" attributes on
typedefs.
As part of this change, I moved the GNU-specific logic (such as
determining the alignment of void or of a function pointer) out of the
expression evaluator and into ASTContext::getTypeInfo.
llvm-svn: 70497
compatible with VC++ and GCC. The codegen/mangling angle hasn't
been fully ironed out yet. Note that we accept int128_t even in
32-bit mode, unlike gcc.
llvm-svn: 70464
SEL, Class, Protocol, CFConstantString, and
__objcFastEnumerationState. With this, we can now run the Objective-C
methods and properties PCH tests.
llvm-svn: 69932
methods, class methods, and property implementations) and instead
place all of these entities into the DeclContext.
This eliminates more linear walks when looking for class or instance
methods and should make PCH (de-)serialization of ObjCDecls trivial
(and lazy).
llvm-svn: 69849
PCH files now contain complete information about builtins, including
any declarations that have been synthesized as part of building the
PCH file. When using a PCH file, we do not initialize builtins at all;
when needed, they'll be found in the PCH file.
This optimization translations into a 9% speedup for "Hello, World!"
with Carbon.h as a prefix header and roughly a 5% speedup for 403.gcc
with its prefix header. We're also reading less of the PCH file for
"Hello, World!":
*** PCH Statistics:
286/20693 types read (1.382110%)
1630/59230 declarations read (2.751984%)
764/44914 identifiers read (1.701029%)
1/32954 statements read (0.003035%)
5/6187 macros read (0.080815%)
down from
*** PCH Statistics:
411/20693 types read (1.986179%)
2553/59230 declarations read (4.310316%)
1093/44646 identifiers read (2.448148%)
1/32954 statements read (0.003035%)
21/6187 macros read (0.339421%)
llvm-svn: 69815
Rework the shadow struct that is layed out for Objective-C classes.
- Superclasses are now always laid out in their shadow structure at
the first field.
- Prior to this, the entire class heirarchy was flattened into a
single structure which meant that alignment, padding, and bitfields
were incorrect (the ASTRecordLayout was correct however, which
meant our debug info didn't coincide with ivar offsets, for
example).
- This is still very suboptimal (for example, ivar are looked up
recursively, but I believe the ivar layout itself is now at least
close to correct.
- <rdar://problem/6773388> error: objc[29823]: layout bitmap sliding
backwards
llvm-svn: 69811
- Superclasses are now always laid out their shadow structure at the
first field.
- Prior to this, the entire class heirarchy was flattened into a
single structure which meant that alignment, padding, and bitfields
weren't packed correctly (the ASTRecordLayout was correct however,
which meant our debug info didn't coincide with ivar offsets, for
example).
- This is still very suboptimal, but I believe the ivar layout itself
is now at least close to correct.
- <rdar://problem/6773388> error: objc[29823]: layout bitmap sliding
backwards
llvm-svn: 69771
@implementation that closes a @class delcaration.
- I don't know how to make a test case for this, but this strengthens
the invariants that hold internally. The functionality change here
is the edit to SemaDeclObjC.cpp.
llvm-svn: 69728
This fixes <rdar://problem/6782722> XCDataTipsManager.m registers, observes notifications in class methods.
The radar above is the result of clang typing 'self' in a class method as 'Class', which results in some spurious warnings (GCC types 'self' in a class method as 'id').
I considered changing the type of 'self' to 'id' (to conform to GCC), however this resulted in *many* test cases breaking. In addition, I really prefer a more strongly typed 'self'.
All in all, this is the least obtrusive fix I could find for removing the spurious warnings (though we do loose some valid warnings).
llvm-svn: 69041
de-serialization of abstract syntax trees.
PCH support serializes the contents of the abstract syntax tree (AST)
to a bitstream. When the PCH file is read, declarations are serialized
as-needed. For example, a declaration of a variable "x" will be
deserialized only when its VarDecl can be found by a client, e.g.,
based on name lookup for "x" or traversing the entire contents of the
owner of "x".
This commit provides the framework for serialization and (lazy)
deserialization, along with support for variable and typedef
declarations (along with several kinds of types). More
declarations/types, along with important auxiliary structures (source
manager, preprocessor, etc.), will follow.
llvm-svn: 68732
types. It is no longer needed now that the code generator
re-lays-out interfaces if they are defines after being laid out
from a forward decl.
llvm-svn: 68194
In a case like:
@class foo;
foo *P;
addRecordToClass was making an empty shadow struct for the foo interface and
completing it. Later when an:
@interface foo
...
@endif
foo *Q;
was seen, ASTContext::addRecordToClass would think that foo was already laid
out and not lay out the definition. This fixes it to create a forward declared
struct the first time around, then complete it when the definition is seen.
Note that this causes two tests to regress, because something is trying to get
the size of the forward declared structs returned by this. Previously, this
would end up getting a size of zero but now it properly dies. I'm not sure
what the right solution is for this, so I xfailed the tests.
Fariborz, please take a look at this. The testcase in rdar://6676794 now gets
farther, but dies later because the objc ivar is not assigned a field number.
As an aside, I really don't like the fact that the objc front-end is creating
shadow C structs for ObjC types. This seems like an implementation detail of
the code generator that could be fixed by better factoring of the extant code.
llvm-svn: 68106
within nested-name-specifiers, e.g., for the "apply" in
typename MetaFun::template apply<T1, T2>::type
At present, we can't instantiate these nested-name-specifiers, so our
testing is sketchy.
llvm-svn: 68081
representation handles the various ways in which one can name a
template, including unqualified references ("vector"), qualified
references ("std::vector"), and dependent template names
("MetaFun::template apply").
One immediate effect of this change is that the representation of
nested-name-specifiers in type names for class template
specializations (e.g., std::vector<int>) is more accurate. Rather than
representing std::vector<int> as
std::(vector<int>)
we represent it as
(std::vector)<int>
which more closely follows the C++ grammar.
Additionally, templates are no longer represented as declarations
(DeclPtrTy) in Parse-Sema interactions. Instead, I've introduced a new
OpaquePtr type (TemplateTy) that holds the representation of a
TemplateName. This will simplify the handling of dependent
template-names, once we get there.
llvm-svn: 68074
instantiation for C++ typename-specifiers such as
typename T::type
The parsing of typename-specifiers is relatively easy thanks to
annotation tokens. When we see the "typename", we parse the
typename-specifier and produce a typename annotation token. There are
only a few places where we need to handle this. We currently parse the
typename-specifier form that terminates in an identifier, but not the
simple-template-id form, e.g.,
typename T::template apply<U, V>
Parsing of nested-name-specifiers has a similar problem, since at this
point we don't have any representation of a class template
specialization whose template-name is unknown.
Semantic analysis is only partially complete, with some support for
template instantiation that works for simple examples.
llvm-svn: 67875
uniqued representation that should both save some memory and make it
far easier to properly build canonical types for types involving
dependent nested-name-specifiers, e.g., "typename T::Nested::type".
This approach will greatly simplify the representation of
CXXScopeSpec. That'll be next.
llvm-svn: 67799
isObjCObjectPointerType to work with qualified types. Adjust test for
changes.
If the SemaExpr changes are wrong or break existing code, feel free to
delete the "ExprTy.addConst();" line and revert my changes to
test/Sema/block-literal.c.
llvm-svn: 67489
qualified name, e.g.,
foo::x
so that we retain the nested-name-specifier as written in the source
code and can reproduce that qualified name when printing the types
back (e.g., in diagnostics). This is PR3493, which won't be complete
until finished the other tasks mentioned near the end of this commit.
The parser's representation of nested-name-specifiers, CXXScopeSpec,
is now a bit fatter, because it needs to contain the scopes that
precede each '::' and keep track of whether the global scoping
operator '::' was at the beginning. For example, we need to keep track
of the leading '::', 'foo', and 'bar' in
::foo::bar::x
The Action's CXXScopeTy * is no longer a DeclContext *. It's now the
opaque version of the new NestedNameSpecifier, which contains a single
component of a nested-name-specifier (either a DeclContext * or a Type
*, bitmangled).
The new sugar type QualifiedNameType composes a sequence of
NestedNameSpecifiers with a representation of the type we're actually
referring to. At present, we only build QualifiedNameType nodes within
Sema::getTypeName. This will be extended to other type-constructing
actions (e.g., ActOnClassTemplateId).
Also on the way: QualifiedDeclRefExprs will also store a sequence of
NestedNameSpecifiers, so that we can print out the property
nested-name-specifier. I expect to also use this for handling
dependent names like Fibonacci<I - 1>::value.
llvm-svn: 67265
such as replacing 'T' in vector<T>. There are a few aspects to this:
- Extend TemplateArgument to allow arbitrary expressions (an
Expr*), and switch ClassTemplateSpecializationType to store
TemplateArguments rather than it's own type-or-expression
representation.
- ClassTemplateSpecializationType can now store dependent types. In
that case, the canonical type is another
ClassTemplateSpecializationType (with default template arguments
expanded) rather than a declaration (we don't build Decls for
dependent types).
- Split ActOnClassTemplateId into ActOnClassTemplateId (called from
the parser) and CheckClassTemplateId (called from
ActOnClassTemplateId and InstantiateType). They're smart enough to
handle dependent types, now.
llvm-svn: 66509
giving them rough classifications (normal types, never-canonical
types, always-dependent types, abstract type representations) and
making it far easier to make sure that we've hit all of the cases when
decoding types.
Switched some switch() statements on the type class over to using this
mechanism, and filtering out those things we don't care about. For
example, CodeGen should never see always-dependent or non-canonical
types, while debug info generation should never see always-dependent
types. More switch() statements on the type class need to be moved
over to using this approach, so that we'll get warnings when we add a
new type then fail to account for it somewhere in the compiler.
As part of this, some types have been renamed:
TypeOfExpr -> TypeOfExprType
FunctionTypeProto -> FunctionProtoType
FunctionTypeNoProto -> FunctionNoProtoType
There shouldn't be any functionality change...
llvm-svn: 65591
nicely sugared type that shows how the user wrote the actual
specialization. This sugared type won't actually show up until we
start doing instantiations.
llvm-svn: 65577
I know, these follow the exact same rules as pointers, so I just made
them use the same codepath. Someone more familiar with ObjC should
double-check this, though.
llvm-svn: 65261
Found while researching <rdar://problem/6497631> Message lookup is sometimes different than gcc's.
Will never be seen in user code. Needed to pass dejagnu testsuite.
llvm-svn: 65244
Should clang have a config.h or should we use the config.h of llvm or using the preprocessor is OK? I did a quick fix here, but having a guideline on how to handle non portable function would be great (or ask ted to stop breaking the windows build :)).
llvm-svn: 65233
Move two key ObjC typechecks from Sema::CheckPointerTypesForAssignment() to ASTContext::mergeTypes().
This allows us to take advantage of the recursion in ASTContext::mergeTypes(), removing some bogus warnings.
This test case I've added includes an example where we still warn (and GCC doesn't). Need to talk with folks and decide what to do. At this point, the major bogosities should be fixed.
llvm-svn: 65231
- Renamed to getDeclAlignInBytes since most other query functions
work in bits.
- Fun to track down as isIntegerConstantExpr was getting it right,
but Evaluate() was getting it wrong. Maybe we should assert they
compute the same thing when they succeed?
llvm-svn: 64828
which consequently caused a Seg fault. during meta-data
generation. It also addresses an issue related to
late binding of newly synthesize ivars (when we support it).
llvm-svn: 64563
Currently only used for 128-bit integers.
Note that we can't use the fixed-width integer types for other integer
modes without other changes because glibc headers redefines (u)int*_t
and friends using the mode attribute. For example, this means that uint64_t
has to be compatible with unsigned __attribute((mode(DI))), and
uint64_t is currently defined to long long. And I have a feeling we'll
run into issues if we try to define uint64_t as something which isn't
either long or long long.
This doesn't get the alignment right in most cases, including
the 128-bit integer case; I'll file a PR shortly. The gist of the issue
is that the targets don't really expose the information necessary to
figure out the alignment outside of the target description, so there's a
non-trivial amount of work involved in getting it working right. That
said, the alignment used is conservative, so the only issue with the
current implementation is ABI compatibility.
This makes it trivial to add some sort of "bitwidth" attribute to make
arbitrary-width integers; I'll do that in a followup.
We could also use this for stuff like the following for compatibility
with gcc, but I have a feeling it would be a better idea for clang to be
consistent between C and C++ modes rather than follow gcc's example for
C mode.
struct {unsigned long long x : 33;} x;
unsigned long long a(void) {return x.x+1;}
llvm-svn: 64434
- rename isObjCIdType/isObjCClassType -> isObjCIdStructType/isObjCClassStructType. The previous name didn't do what you would expect.
- add back isObjCIdType/isObjCClassType to do what you would expect. Not currently used, however many of the isObjCIdStructType/isObjCClassStructType clients could be converted over time.
- move static Sema function areComparableObjCInterfaces to ASTContext (renamed to areComparableObjCPointerTypes, since it now operates on pointer types).
llvm-svn: 64385
to a class template. For example, the template-id 'vector<int>' now
has a nice, sugary type in the type system. What we can do now:
- Parse template-ids like 'vector<int>' (where 'vector' names a
class template) and form proper types for them in the type system.
- Parse icky template-ids like 'A<5>' and 'A<(5 > 0)>' properly,
using (sadly) a bool in the parser to tell it whether '>' should
be treated as an operator or not.
This is a baby-step, with major problems and limitations:
- There are currently two ways that we handle template arguments
(whether they are types or expressions). These will be merged, and,
most likely, TemplateArg will disappear.
- We don't have any notion of the declaration of class template
specializations or of template instantiations, so all template-ids
are fancy names for 'int' :)
llvm-svn: 64153
canonicalize by template parameter depth, index, and name, and the
unnamed version of a template parameter serves as the canonical.
TemplateTypeParmDecl no longer needs to inherit from
TemplateParmPosition, since depth and index information is present
within the type.
llvm-svn: 63899
Also changed FunctionTypeProto to be allocated with 8-byte alignment (noticed by Doug). I couldn't think of any reason to allocate on 16-byte boundaries. If anyone remembers why we were doing this, let me know!
llvm-svn: 63137
that every declaration lives inside a DeclContext.
Moved several things that don't have names but were ScopedDecls (and,
therefore, NamedDecls) to inherit from Decl rather than NamedDecl,
including ObjCImplementationDecl and LinkageSpecDecl. Now, we don't
store empty DeclarationNames for these things, nor do we try to insert
them into DeclContext's lookup structure.
The serialization tests are temporarily disabled. We'll re-enable them
once we've sorted out the remaining ownership/serialiazation issues
between DeclContexts and TranslationUnion, DeclGroups, etc.
llvm-svn: 62562
even when we are still defining the TagDecl. This is required so that
qualified name lookup of a class name within its definition works (see
the new bits in test/SemaCXX/qualified-id-lookup.cpp).
As part of this, move the nested redefinition checking code into
ActOnTag. This gives us diagnostics earlier (when we try to perform
the nested redefinition, rather than when we try to complete the 2nd
definition) and removes some code duplication.
llvm-svn: 62386
filters the decls seen by decl_iterator with two criteria: the dynamic
type of the declaration and a run-time predicate described by a member
function. This simplifies EnumDecl, RecordDecl, and ObjCContainerDecl
considerably. It has no measurable performance impact.
llvm-svn: 61994
Douglas Gregor