return value optimization. Sema marks return statements with their
NRVO candidates (which may or may not end up using the NRVO), then, at
the end of a function body, computes and marks those variables that
can be allocated into the return slot.
I've checked this locally with some debugging statements (not
committed), but there won't be any tests until CodeGen comes along.
llvm-svn: 103865
user directive is needed to force a property implementation.
It is decided based on those propeties which are declared in
the class (or in its protocols) but not those which must be
default implemented by one of its super classes. Implements radar 7923851.
llvm-svn: 103787
"used" (e.g., we will refer to the vtable in the generated code) and
when they are defined (i.e., because we've seen the key function
definition). Previously, we were effectively tracking "potential
definitions" rather than uses, so we were a bit too eager about emitting
vtables for classes without key functions.
The new scheme:
- For every use of a vtable, Sema calls MarkVTableUsed() to indicate
the use. For example, this occurs when calling a virtual member
function of the class, defining a constructor of that class type,
dynamic_cast'ing from that type to a derived class, casting
to/through a virtual base class, etc.
- For every definition of a vtable, Sema calls MarkVTableUsed() to
indicate the definition. This happens at the end of the translation
unit for classes whose key function has been defined (so we can
delay computation of the key function; see PR6564), and will also
occur with explicit template instantiation definitions.
- For every vtable defined/used, we mark all of the virtual member
functions of that vtable as defined/used, unless we know that the key
function is in another translation unit. This instantiates virtual
member functions when needed.
- At the end of the translation unit, Sema tells CodeGen (via the
ASTConsumer) which vtables must be defined (CodeGen will define
them) and which may be used (for which CodeGen will define the
vtables lazily).
From a language perspective, both the old and the new schemes are
permissible: we're allowed to instantiate virtual member functions
whenever we want per the standard. However, all other C++ compilers
were more lazy than we were, and our eagerness was both a performance
issue (we instantiated too much) and a portability problem (we broke
Boost test cases, which now pass).
Notes:
(1) There's a ton of churn in the tests, because the order in which
vtables get emitted to IR has changed. I've tried to isolate some of
the larger tests from these issues.
(2) Some diagnostics related to
implicitly-instantiated/implicitly-defined virtual member functions
have moved to the point of first use/definition. It's better this
way.
(3) I could use a review of the places where we MarkVTableUsed, to
see if I missed any place where the language effectively requires a
vtable.
Fixes PR7114 and PR6564.
llvm-svn: 103718
member function (default constructor, copy constructor, copy
assignment operator, destructor), emit a note showing where that
implicit definition was required.
llvm-svn: 103619
referenced unless we see one of them defined (or the key function
defined, if it as one) or if we need the vtable for something. Fixes
PR7114.
llvm-svn: 103497
particular, don't complain about unused variables that have dependent
type until instantiation time, so that we can look at the type of the
variable. Moreover, only complain about unused variables that have
neither a user-declared constructor nor a non-trivial destructor.
llvm-svn: 103362
if/switch/while/do/for statements. Previously, we would end up either:
(1) Forgetting to destroy temporaries created in the condition (!),
(2) Destroying the temporaries created in the condition *before*
converting the condition to a boolean value (or, in the case of a
switch statement, to an integral or enumeral value), or
(3) In a for statement, destroying the condition's temporaries at
the end of the increment expression (!).
We now destroy temporaries in conditions at the right times. This
required some tweaking of the Parse/Sema interaction, since the parser
was building full expressions too early in many places.
Fixes PR7067.
llvm-svn: 103187
"bottom-up" when implicit casts and comparisons are inserted, compute them
"top-down" when the full expression is finished. Makes it easier to
coordinate warnings and thus implement -Wconversion for signedness
conversions without double-warning with -Wsign-compare. Also makes it possible
to realize that a signedness conversion is okay because the context is
performing the inverse conversion. Also simplifies some logic that was
trying to calculate the ultimate comparison/result type and getting it wrong.
Also fixes a problem with the C++ explicit casts which are often "implemented"
in the AST with a series of implicit cast expressions.
llvm-svn: 103174
friend function template, be sure to adjust the computed template
argument lists based on the location of the definition of the function
template: it's possible that the definition we're instantiating with
and the template declaration that we found when creating the
specialization are in different contexts, which meant that we would
end up using the wrong template arguments for instantiation.
Fixes PR7013; all Boost.DynamicBitset tests now pass.
llvm-svn: 102974
conforms to a protocol as one of its super classes does. This is because
conforming super class will implement the property. This implements
new warning rules for unimplemented properties (radar 7884086).
llvm-svn: 102919
assignment operators.
Previously, Sema provided type-checking and template instantiation for
copy assignment operators, then CodeGen would synthesize the actual
body of the copy constructor. Unfortunately, the two were not in sync,
and CodeGen might pick a copy-assignment operator that is different
from what Sema chose, leading to strange failures, e.g., link-time
failures when CodeGen called a copy-assignment operator that was not
instantiation, run-time failures when copy-assignment operators were
overloaded for const/non-const references and the wrong one was
picked, and run-time failures when by-value copy-assignment operators
did not have their arguments properly copy-initialized.
This implementation synthesizes the implicitly-defined copy assignment
operator bodies in Sema, so that the resulting ASTs encode exactly
what CodeGen needs to do; there is no longer any special code in
CodeGen to synthesize copy-assignment operators. The synthesis of the
body is relatively simple, and we generate one of three different
kinds of copy statements for each base or member:
- For a class subobject, call the appropriate copy-assignment
operator, after overload resolution has determined what that is.
- For an array of scalar types or an array of class types that have
trivial copy assignment operators, construct a call to
__builtin_memcpy.
- For an array of class types with non-trivial copy assignment
operators, synthesize a (possibly nested!) for loop whose inner
statement calls the copy constructor.
- For a scalar type, use built-in assignment.
This patch fixes at least a few tests cases in Boost.Spirit that were
failing because CodeGen picked the wrong copy-assignment operator
(leading to link-time failures), and I suspect a number of undiagnosed
problems will also go away with this change.
Some of the diagnostics we had previously have gotten worse with this
change, since we're going through generic code for our
type-checking. I will improve this in a subsequent patch.
llvm-svn: 102853
information required to implicitly define a C++ special member
function. Use it rather than explicitly setting CurContext on entry
and exit, which is fragile.
Use this RAII object for the implicitly-defined default constructor,
copy constructor, copy assignment operator, and destructor.
llvm-svn: 102840
of the mapping from local declarations to their instantiated
counterparts during template instantiation. Previously, we tried to do
some unholy merging of local instantiation scopes that involved
storing a single hash table along with an "undo" list on the
side... which was ugly, and never handled function parameters
properly.
Now, we just keep separate hash tables for each local instantiation
scope, and "combining" two scopes means that we'll look in each of the
combined hash tables. The combined scope stack is rarely deep, and
this makes it easy to avoid the "undo" issues we were hitting. Also,
I've simplified the logic for function parameters: if we're declaring
a function and we need the function parameters to live longer, we just
push them back into the local instantiation scope where we need them.
Fixes PR6990.
llvm-svn: 102732
entering the current instantiation. Set up a little to preserve type location
information for typename types while we're in there.
Fixes a Boost failure.
llvm-svn: 102673
template argument deduction or (more importantly) the final substitution
required by such deduction. Makes access control magically work in these
cases.
Fixes PR6967.
llvm-svn: 102572
Amadini.
This change introduces a new expression node type, OffsetOfExpr, that
describes __builtin_offsetof. Previously, __builtin_offsetof was
implemented using a unary operator whose subexpression involved
various synthesized array-subscript and member-reference expressions,
which was ugly and made it very hard to instantiate as a
template. OffsetOfExpr represents the AST more faithfully, with proper
type source information and a more compact representation.
OffsetOfExpr also has support for dependent __builtin_offsetof
expressions; it can be value-dependent, but will never be
type-dependent (like sizeof or alignof). This commit introduces
template instantiation for __builtin_offsetof as well.
There are two major caveats to this patch:
1) CodeGen cannot handle the case where __builtin_offsetof is not a
constant expression, so it produces an error. So, to avoid
regressing in C, we retain the old UnaryOperator-based
__builtin_offsetof implementation in C while using the shiny new
OffsetOfExpr implementation in C++. The old implementation can go
away once we have proper CodeGen support for this case, which we
expect won't cause much trouble in C++.
2) __builtin_offsetof doesn't work well with non-POD class types,
particularly when the designated field is found within a base
class. I will address this in a subsequent patch.
Fixes PR5880 and a bunch of assertions when building Boost.Python
tests.
llvm-svn: 102542
of a class template or class template partial specialization. That is to
say, in
template <class T> class A { ... };
or
template <class T> class B<const T*> { ... };
make 'A<T>' and 'B<const T*>' sugar for the corresponding InjectedClassNameType
when written inside the appropriate context. This allows us to track the
current instantiation appropriately even inside AST routines. It also allows
us to compute a DeclContext for a type much more efficiently, at some extra
cost every time we write a template specialization (which can be optimized,
but I've left it simple in this patch).
llvm-svn: 102407
by using TypeSourceInfo, cleaning up the representation
somewhat. Teach getTypeOperand() to strip references and
cv-qualifiers, providing the semantic view of the type without
requiring any extra storage (the unmodified type remains within the
TypeSourceInfo). This fixes a bug found by Boost's call_traits test.
Finally, clean up semantic analysis, by splitting the ActOnCXXTypeid
routine into ActOnCXXTypeId (the parser action) and two BuildCXXTypeId
functions, which perform the semantic analysis for typeid(type) and
typeid(expression), respectively. We now perform less work at template
instantiation time (we don't look for std::type_info again) and can
give better diagnostics.
llvm-svn: 102393
function-parameter checking and splitting it into the normal
ActOn*/Build* pair in Sema. We now use VarDecl to represent the @catch
parameter rather than the ill-fitting ParmVarDecl.
llvm-svn: 102347
(e.g., no typename, enum, class, etc.), e.g., because the context is
one that is known to refer to a type. Patch from Enea Zaffanella!
llvm-svn: 102243
arguments. Rather than having the parser call ActOnParamDeclarator
(which is a bit of a hack), call a new ActOnObjCExceptionDecl
action. We'll be moving more functionality into this handler to
perform earlier checking of @catch.
llvm-svn: 102222
statements. Instead of the @try having a single @catch, where all of
the @catch's were chained (using an O(n^2) algorithm nonetheless),
@try just holds an array of its @catch blocks. The resulting AST is
slightly more compact (not important) and better represents the actual
language semantics (good).
llvm-svn: 102221
method being called at template definition time, retain that method
and pass it through to type-checking. We will not perform any lookup
for the method during template instantiation.
llvm-svn: 102081
method parameter, provide a note pointing at the parameter itself so
the user does not have to manually look for the function/method being
called and match up parameters to arguments. For example, we now get:
t.c:4:5: warning: incompatible pointer types passing 'long *' to
parameter of
type 'int *' [-pedantic]
f(long_ptr);
^~~~~~~~
t.c:1:13: note: passing argument to parameter 'x' here
void f(int *x);
^
llvm-svn: 102038
Objective-C class message expression into a type from the parser
(which was doing so in two places) to Action::getObjCMessageKind()
which, in the case of Sema, reduces the number of name lookups we need
to perform.
llvm-svn: 102026
sends. Major changes include:
- Expanded the interface from two actions (ActOnInstanceMessage,
ActOnClassMessage), where ActOnClassMessage also handled sends to
"super" by checking whether the identifier was "super", to three
actions (ActOnInstanceMessage, ActOnClassMessage,
ActOnSuperMessage). Code completion has the same changes.
- The parser now resolves the type to which we are sending a class
message, so ActOnClassMessage now accepts a TypeTy* (rather than
an IdentifierInfo *). This opens the door to more interesting
types (for Objective-C++ support).
- Split ActOnInstanceMessage and ActOnClassMessage into parser
action functions (with their original names) and semantic
functions (BuildInstanceMessage and BuildClassMessage,
respectively). At present, this split is onyl used by
ActOnSuperMessage, which decides which kind of super message it
has and forwards to the appropriate Build*Message. In the future,
Build*Message will be used by template instantiation.
- Use getObjCMessageKind() within the disambiguation of Objective-C
message sends vs. array designators.
Two notes about substandard bits in this patch:
- There is some redundancy in the code in ParseObjCMessageExpr and
ParseInitializerWithPotentialDesignator; this will be addressed
shortly by centralizing the mapping from identifiers to type names
for the message receiver.
- There is some #if 0'd code that won't likely ever be used---it
handles the use of 'super' in methods whose class does not have a
superclass---but could be used to model GCC's behavior more
closely. This code will die in my next check-in, but I want it in
Subversion.
llvm-svn: 102021
Douglas Gregor