The approach I've taken in this patch is relatively straightforward,
although the code itself is non-trivial. Essentially, as we process
an initializer list we build up a fully-explicit representation of the
initializer list, where each of the subobject initializations occurs
in order. Designators serve to "fill in" subobject initializations in
a non-linear way. The fully-explicit representation makes initializer
lists (both with and without designators) easy to grok for codegen and
later semantic analyses. We keep the syntactic form of the initializer
list linked into the AST for those clients interested in exactly what
the user wrote.
Known limitations:
- Designating a member of a union that isn't the first member may
result in bogus initialization (we warn about this)
- GNU array-range designators are not supported (we warn about this)
llvm-svn: 63242
accurately states what the function is trying to do and how it is
different from Expr::isEvaluatable. Also get rid of a parameter that is both
unused and inaccurate.
llvm-svn: 62951
and separates lexical name lookup from qualified name lookup. In
particular:
* Make DeclContext the central data structure for storing and
looking up declarations within existing declarations, e.g., members
of structs/unions/classes, enumerators in C++0x enums, members of
C++ namespaces, and (later) members of Objective-C
interfaces/implementations. DeclContext uses a lazily-constructed
data structure optimized for fast lookup (array for small contexts,
hash table for larger contexts).
* Implement C++ qualified name lookup in terms of lookup into
DeclContext.
* Implement C++ unqualified name lookup in terms of
qualified+unqualified name lookup (since unqualified lookup is not
purely lexical in C++!)
* Limit the use of the chains of declarations stored in
IdentifierInfo to those names declared lexically.
* Eliminate CXXFieldDecl, collapsing its behavior into
FieldDecl. (FieldDecl is now a ScopedDecl).
* Make RecordDecl into a DeclContext and eliminates its
Members/NumMembers fields (since one can just iterate through the
DeclContext to get the fields).
llvm-svn: 60878
initializers. llvm-gcc appears to be more aggressive, but incorrect,
for constructs like "const int a[] = {1,2,3};"; that said, current
optimizers will do the appropriate optimizations when safe.
llvm-svn: 60270
- Use dotted notation for blocks related to a particular statement
type.
- Use .end for landing pads.
No functionality change in NDEBUG mode. :)
llvm-svn: 59210
- Change Obj-C runtime message API, drop the ObjCMessageExpr arg in
favor of just result type and selector. Necessary so it can be
reused in situations where we don't want to cons up an
ObjCMessageExpr.
- Update aggregate binary assignment to know about special property
ref lvalues.
- Add CodeGenFunction::EmitCallArg overload which takes an already
emitted rvalue.
Add CodeGenFunction::StoreComplexIntoAddr.
Disabled logic in Sema for parsing Objective-C dot-syntax that
accesses methods. This code does not search in the correct order and
the AST node has no way of properly representing its results.
Updated StmtDumper to print a bit more information about
ObjCPropertyRefExprs.
llvm-svn: 55561
Implement Obj-C lvalue message sends (aggregate returns).
Update several places to emit more precise ErrorUnsupported warnings
for currently unimplemented Obj-C features (main missing chunks are
property references, Obj-C exception handling, and the for ... in
syntax).
llvm-svn: 55234
- Returns an RValue.
- Reduced to only taking the CodeGenFunction, Expr, and Receiver.
- Becomes responsible for emitting the arguments.
Add CodeGenFunction::EmitCallExprExt
- Takes optional extra arguments to insert at the head of the call.
- This allows the Obj-C runtimes to call into this and isolates the
argument and call instruction generation code to one place. Upshot
is that we now pass structures (more) correctly.
Also, fix one aspect of generating methods which take structure
arguments (for NeXT). This probably needs to be merged with the
SetFunctionAttributes code in CodeGenModule.cpp
llvm-svn: 55223
move getAsArrayType into ASTContext instead of being a method on type.
This is required because getAsArrayType(const AT), where AT is a typedef
for "int[10]" needs to return ArrayType(const int, 10).
Fixing this greatly simplifies getArrayDecayedType, which is a good sign.
llvm-svn: 54317
clang as a Release build.
The big change is that all AST nodes (subclasses of Stmt) whose children are
Expr* store their children as Stmt* or arrays of Stmt*. This is to remove
strict-aliasing warnings when using StmtIterator. None of the interfaces of any
of the classes have changed (except those with arg_iterators, see below), as the
accessor methods introduce the needed casts (via cast<>). While this extra
casting may seem cumbersome, it actually adds some important sanity checks
throughout the codebase, as clients using StmtIterator can potentially overwrite
children that are expected to be Expr* with Stmt* (that aren't Expr*). The casts
provide extra sanity checks that are operational in debug builds to catch
invariant violations such as these.
For classes that have arg_iterators (e.g., CallExpr), the definition of
arg_iterator has been replaced. Instead of it being Expr**, it is an actual
class (called ExprIterator) that wraps a Stmt**, and provides the necessary
operators for iteration. The nice thing about this class is that it also uses
cast<> to type-checking, which introduces extra sanity checks throughout the
codebase that are useful for debugging.
A few of the CodeGen functions that use arg_iterator (especially from
OverloadExpr) have been modified to take begin and end iterators instead of a
base Expr** and the number of arguments. This matches more with the abstraction
of iteration. This still needs to be cleaned up a little bit, as clients expect
that ExprIterator is a RandomAccessIterator (which we may or may not wish to
allow for efficiency of representation).
This is a fairly large patch. It passes the tests (except CodeGen/bitfield.c,
which was already broken) on both a Debug and Release build, but it should
obviously be reviewed.
llvm-svn: 52378
qualifier in the lvalue, and changes lvalue loads/stores to honor
the volatile flag. Places which need some further attention are marked
with FIXMEs.
Patch by Cédric Venet.
llvm-svn: 52264
required for correctness in cases of copying a struct to itself or to
an overlapping struct (itself for cases like *a = *a, and overlapping
is possible with unions).
Hopefully, this won't end up being a perf issue; LLVM *should* be able
to optimize memmove to memcpy in a lot of cases, and for small copies
the generated code *should* be mostly comparable. (In reality, LLVM
is currently horrible at optimizing memmove, but that's a bug, not a
fundamental issue.)
gcc currently generates wrong code; that's
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32667.
llvm-svn: 51566
llvm::Type::isSingleValueType. Currently these two functions have
the same behavior, but soon isFirstClassType will return true for
struct and array types.
Clang may some day want to use of isFirstClassType for some of
these some day as an optimization, but it'll require some
consideration.
llvm-svn: 51446
they were causing bad code to be emitted. There are two fixes here: one
makes sure we emit a string that is long enough, and one makes sure we
properly handle string initialization in init lists.
llvm-svn: 51259
Douglas Gregor