two classes, one for typenames and one for values; this seems to have some
support from Doug if not necessarily from the extremely-vague-on-this-point
standard. Track the location of the 'typename' keyword in a using-typename
decl. Make a new lookup result for unresolved values and deal with it in
most places.
llvm-svn: 89184
that we're dealing with canonical types like the documentation say
(yay, CanQualType). Alas, this is another instance where using
getQualifiers() on a non-canonical QualType got us in trouble.
Good news: with this fix, Clang can now parse all of its own headers!
llvm-svn: 88848
set, expand overloaded function declarations. Long-term, this should
actually be done by the name-lookup code rather than here, but this
part of the code (involving using declarations) is getting a makeover
now and the test-case is useful.
llvm-svn: 88846
type, use full qualified name lookup rather than the poking the
declaration context directly. This makes sure that we see operator()'s
in superclasses. Also, move the complete-type check before this name
lookup.
llvm-svn: 88842
- Also, perform calculated implicit cast sequences if they're determined to work. This finally diagnoses static_cast to ambiguous or implicit bases and fixes two long-standing fixmes in the test case. For the C-style cast, this requires propagating the access check suppression pretty deep into other functions.
- Pass the expressions for TryStaticCast and TryStaticImplicitCast by reference. This should lead to a better AST being emitted for such casts, and also fixes a memory leak, because CheckReferenceInit and PerformImplicitConversion wrap the node passed to them. These wrappers were previously lost.
llvm-svn: 88809
cast) that is converting to a class type, enumerate its constructors
as in any other direct initialization. This ensures that we get the
proper conversion sequence.
llvm-svn: 88751
a class type from itself or a derived class thereof, enumerate
constructors and permit user-defined conversions to the arguments of
those constructors. This fixes the wacky implicit conversion sequence
used in std::auto_ptr's lame emulation of move semantics.
llvm-svn: 88670
and implicitly defined constructors. This has a number of benefits:
1. Less code.
2. Explicit and implicit constructors get the same diagnostics.
3. The AST explicitly contains constructor calls from implicit default
constructors. This allows handing some cases that previously weren't handled
correctly in IRGen without any additional code. Specifically, implicit default
constructors containing calls to constructors with default arguments are now
handled correctly.
llvm-svn: 86500
* If the unsigned type is smaller than the signed type, never warn, because
its value will not change when zero-extended to the larger type.
* If we're testing for (in)equality, and the unsigned value is an integer
constant whose sign bit is not set, never warn, because even though the
signed value might change, it can't affect the result of the equality.
Also make the comparison test cases much more rigorous, and have them expose
the subtle differences between C and C++ here.
llvm-svn: 86242
get_origin->x
where get_origin is actually a function and the user has forgotten the
parentheses. Instead of giving a lame note for the fix-it, give a
full-fledge error, early, then build the call expression to try to
recover.
llvm-svn: 86238
class type, don't perform the array-to-pointer or function-to-pointer
conversions, because we may end up binding a reference to a function
or array.
With this change, FileCheck now passes -fsyntax-only!
llvm-svn: 86211
always zero in this context" warning logic. Also, make the diagnostic
itself more precise when referring to pointer values ("NULL" vs. "zero").
llvm-svn: 86143
DiagnoseSignCompare into Sema::CheckSignCompare and call it from more places.
Add some enumerator tests. These seem to expose some oddities in the
types we're converting C++ enumerators to; in particular, they're converting
to unsigned before int, which seems to contradict 4.5 [conv.prom] p2.
Note to self: stop baiting Doug in my commit messages.
llvm-svn: 86128
"->" with a use of ParseUnqualifiedId. Collapse
ActOnMemberReferenceExpr, ActOnDestructorReferenceExpr (both of them),
ActOnOverloadedOperatorReferenceExpr,
ActOnConversionOperatorReferenceExpr, and
ActOnMemberTemplateIdReferenceExpr into a single, new action
ActOnMemberAccessExpr that does the same thing more cleanly (and can
keep more source-location information).
llvm-svn: 85930
representation of a C++ unqualified-id, along with a single parsing
function (Parser::ParseUnqualifiedId) that will parse all of the
various forms of unqualified-id in C++.
Replace the representation of the declarator name in Declarator with
the new UnqualifiedId class, simplifying declarator-id parsing
considerably and providing more source-location information to
Sema. In the future, I hope to migrate all of the other
unqualified-id-parsing code over to this single representation, then
begin to merge actions that are currently only different because we
didn't have a unqualified notion of the name in the parser.
llvm-svn: 85851
the DeclaratorInfo, one for semantic analysis), just build a single type whose
canonical type will reflect the semantic analysis (assuming the type is
well-formed, of course).
To make that work, make a few changes to the type system:
* allow the nominal pointee type of a reference type to be a (possibly sugared)
reference type. Also, preserve the original spelling of the reference type.
Both of these can be ignored on canonical reference types.
* Remove ObjCProtocolListType and preserve the associated source information on
the various ObjC TypeLocs. Preserve the spelling of protocol lists except in
the canonical form.
* Preserve some level of source type structure on parameter types, but
canonicalize on the canonical function type. This is still a WIP.
Drops code size, makes strides towards accurate source location representation,
slight (~1.7%) progression on Cocoa.h because of complexity drop.
llvm-svn: 84907
sets of builtin operators. Currently, it is applied
to '++' and '->*' operators. I need to apply it to others
as well. Also, heuristics need be applied to
BuiltinCandidateTypeSet::AddPointerWithMoreQualifiedTypeVariants.
This is WIP.
llvm-svn: 84187
unknown type name, e.g.,
foo::bar x;
when "bar" does not refer to a type in "foo".
With this change, the parser now calls into the action to perform
diagnostics and can try to recover by substituting in an appropriate
type. For example, this allows us to easily diagnose some missing
"typename" specifiers, which we now do:
test/SemaCXX/unknown-type-name.cpp:29:1: error: missing 'typename'
prior to dependent type name 'A<T>::type'
A<T>::type A<T>::f() { return type(); }
^~~~~~~~~~
typename
Fixes PR3990.
llvm-svn: 84053
what we found when we looked into <blah>", where <blah> is a
DeclContext*. We can now format DeclContext*'s in nice ways, e.g.,
"namespace N", "the global namespace", "'class Foo'".
This is part of PR3990, but we're not quite there yet.
llvm-svn: 84028
conditions. Add a fixit to insert the parentheses. Also fix a very minor
possible memory leak in 'for' conditions.
Fixes PR 4876 and rdar://problem/7289172
llvm-svn: 83907
user-defined type conversions, issue list of ambiguites in addition
to the diagnostic. So, clang now issues the following:
b.cpp:19:19: error: left hand operand to ->* must be a pointer to class compatible with the right hand operand, but is 'struct C1'
int i = c1->*pmf;
~~^
b.cpp:19:19: note: because of ambiguity in conversion of 'struct C1' to 'struct E *'
b.cpp:5:5: note: candidate function
operator E*();
^
b.cpp:11:5: note: candidate function
operator E*();
^
llvm-svn: 83862
The exception specification of the assignee must be the same or a subset of the target. In addition, exception specifications on arguments and return types must be equivalent, but this is not implemented yet.
This currently produces two diagnostics for every invalid assignment/initialization, due to the diagnostic produced outside PerformImplicitConversion, e.g. in CheckSingleInitializer. I don't know how to suppress this; in any case I think it is the wrong place for a diagnostic, since there are other diagnostics produced inside the function. So I'm leaving it as it is for the moment.
llvm-svn: 83710
struct B;
B f();
void g() {
f();
}
We now get
t.cpp:6:3: error: calling 'f' with incomplete return type 'struct B'
f();
^~~
t.cpp:3:3: note: 'f' declared here
B f();
^
t.cpp:1:8: note: forward declaration of 'struct B'
struct B;
^
llvm-svn: 83692
Now we produce things like:
bug1.cpp:21:11: error: use of overloaded operator '->*' is ambiguous
int i = c->*pmf; // expected-error {{use of overloaded operator '->*' is ambiguous}} \
~^ ~~~
bug1.cpp:21:11: note: built-in candidate operator ->* ('struct A volatile *', 'int const struct A::*')
bug1.cpp:21:11: note: built-in candidate operator ->* ('struct A volatile *', 'int restrict struct A::*')
...
Still need to look at an issue (indicated as FIXME in the test case).
llvm-svn: 83650
Doug, please review. There is a FIXME in the test case with a question
which is unrelated to this patch (that is, error is issued
before set of builtins are added to the candidate list).
llvm-svn: 83429
of the flow-control checks for falling off the end of a function,
since the return type may instantiate to void. Similarly, if a
return statement has an expression and the return type of the function
is void, don't complain if the expression is type-dependent, since
that type could instantiate to void.
Fixes PR5071.
llvm-svn: 83222
overload candidates (but not the built-in ones). We still rely on the
underlying built-in semantic analysis to produce the initial
diagnostic, then print the candidates following that diagnostic.
One side advantage of this approach is that we can perform more validation
of C++'s operator overloading with built-in candidates vs. the
semantic analysis for those built-in operators: when there are no
viable candidates, we know to expect an error from the built-in
operator handling code. Otherwise, we are not modeling the built-in
semantics properly within operator overloading. This is checked as:
assert(Result.isInvalid() &&
"C++ binary operator overloading is missing
candidates!");
if (Result.isInvalid())
PrintOverloadCandidates(CandidateSet, /*OnlyViable=*/false);
The assert() catches cases where we're wrong in a +Asserts build. The
"if" makes sure that, if this happens in a production clang
(-Asserts), we still build the proper built-in operator and continue
on our merry way. This is effectively what happened before this
change, but we've added the assert() to catch more flies.
llvm-svn: 83175
"usual deallocation function" with two arguments. CodeGen will have to
handle this case specifically, since the value for the second argument
(the size of the allocated object) may have to be computed at run
time.
Fixes the Sema part of PR4782.
llvm-svn: 83080
functions that occur in multiple declaration contexts, e.g., because
some were found via using declarations. Now, isDeclInScope will build
a new overload set (when needed) containing only those declarations
that are actually in scope. This eliminates a problem found with
libstdc++'s <iostream>, where the presence of using
In the longer term, I'd like to eliminate Sema::isDeclInScope in favor
of better handling of the RedeclarationOnly flag in the name-lookup
routines. That way, name lookup only returns the entities that matter,
rather than taking the current two-pass approach of producing too many
results and then filtering our the wrong results. It's not efficient,
and I'm sure that we aren't filtering everywhere we should be.
llvm-svn: 82954
It uses a recent API to find inherited conversion functions to do
the initializer to reference lvalue conversion (and removes a FIXME).
It issues the ambiguity diagnostics when multiple conversions are found.
WIP.
llvm-svn: 82649
to pointer function for delete expression. 2)
Treat type conversion function and its 'const' version
as identical in building the visible conversion list.
llvm-svn: 81930
- Diagnose attempts to add default arguments to templates (or member
functions of templates) after the initial declaration (DR217).
- Improve diagnostics when a default argument is redefined. Now, the
note will always point at the place where the default argument was
previously defined, rather than pointing to the most recent
declaration of the function.
llvm-svn: 81548
integral constant expressions (for conversions to integer types,
naturally). I don't *think* that const_casts will ever get to this
point, but I also can't convince myself that they won't... so I've
taken the safe route and allowed the ICE checking code to look at
const_cast.
llvm-svn: 81453
all of the parent DeclContexts that aren't represented within the
Scope chain. This fixes some name-lookup problems in out-of-line
definitions of members of nested classes.
llvm-svn: 81451
such initializations properly convert constructor arguments and fill
in default arguments where necessary. This also makes the ownership
model more clear.
llvm-svn: 81394
formed without a trailing '(', diagnose the error (these expressions
must be immediately called), emit a fix-it hint, and fix the code.
llvm-svn: 81015
expressions, e.g.,
p->~T()
when p is a pointer to a scalar type.
We don't currently diagnose errors when pseudo-destructor expressions
are used in any way other than by forming a call.
llvm-svn: 81009
The problem this change addresses is that we treat __is_pod and
__is_empty as keywords in C++, because they are built-in type traits
in GCC >= 4.3. However, GNU libstdc++ 4.2 (and possibly earlier
versions) define implementation-detail struct templates named __is_pod
and __is_empty.
This commit solves the problem by recognizing
struct __is_pod
and
struct __is_empty
as special token sequences. When one of these token sequences is
encountered, the keyword (__is_pod or __is_empty) is implicitly
downgraded to an identifier so that parsing can continue. This is an
egregious hack, but it has the virtue of "just working" whether
someone is using libstdc++ 4.2 or not, without the need for special
flags.
llvm-svn: 80988
involve qualified names, e.g., x->Base::f. We now maintain enough
information in the AST to compare the results of the name lookup of
"Base" in the scope of the postfix-expression (determined at template
definition time) and in the type of the object expression.
llvm-svn: 80953
t->Base::f
where t has a dependent type. We save the nested-name-specifier in the
CXXUnresolvedMemberExpr then, during instantiation, substitute into
the nested-name-specifier with the (transformed) object type of t, so
that we get name lookup into the type of the object expression.
Note that we do not yet retain information about name lookup into the
lexical scope of the member access expression, so several regression
tests are still disabled.
llvm-svn: 80925
x->Base::f
We no longer try to "enter" the context of the type that "x" points
to. Instead, we drag that object type through the parser and pass it
into the Sema routines that need to know how to perform lookup within
member access expressions.
We now implement most of the crazy name lookup rules in C++
[basic.lookup.classref] for non-templated code, including performing
lookup both in the context of the type referred to by the member
access and in the scope of the member access itself and then detecting
ambiguities when the two lookups collide (p1 and p4; p3 and p7 are
still TODO). This change also corrects our handling of name lookup
within template arguments of template-ids inside the
nested-name-specifier (p6; we used to look into the scope of the
object expression for them) and fixes PR4703.
I have disabled some tests that involve member access expressions
where the object expression has dependent type, because we don't yet
have the ability to describe dependent nested-name-specifiers starting
with an identifier.
llvm-svn: 80843
pointers, by extending the "composite pointer type" logic to include
member pointer types.
Introduce test cases for member pointer comparisons, including those
that involve the builtin operator candidates implemented earlier.
llvm-svn: 79925
Doug, please look at decltype-crash and instantiate-function-1.mm, I'm not sure
if they are actually testing the right thing / anything.
llvm-svn: 77070
and __has_trivial_constructor builtin pseudo-functions and
additionally implements __has_trivial_copy and __has_trivial_assign,
from John McCall!
llvm-svn: 76916
value. This is on by default, and controlled by -Wreturn-type (-Wmost
-Wall). I believe there should be very few false positives, though
the most interesting case would be:
int() { bar(); }
when bar does:
bar() { while (1) ; }
Here, we assume functions return, unless they are marked with the
noreturn attribute. I can envision a fixit note for functions that
never return normally that don't have a noreturn attribute to add a
noreturn attribute.
If anyone spots other false positives, let me know!
llvm-svn: 76821
1. Make it work correctly with anonymous unions.
2. Don't compute it if the warning isn't enabled.
3. Optimize the algorithm slightly to make it linear time in the
case where we don't produce any warnings.
llvm-svn: 76630
- These kinds of "shotgun" tests are very slow, and do not belong in the
regression suite. If these kinds of tests are regarded to have value, they
should be added to the LLVM test-suite.
- I would actually like to remove all of these tests, but I left Sema/carbon.c
and SemaObjC/cocoa.m...
llvm-svn: 75399
in their order of construction for each class and use it
to to check on propery order of base class construction
under -Wreorder option.
llvm-svn: 75270
these are usually because the parser was thoroughly confused. In addition
to typing the value being declared as an int and hoping for the best, we
mark the value as invalid so we don't get chains of errors when it is
used downstream. In C, implicit int actually is valid, so typing the thing
as int is good and marking it invalid is bad. :)
llvm-svn: 74266
templates.
For example, this now type-checks (but does not instantiate the body
of deref<int>):
template<typename T> T& deref(T* t) { return *t; }
void test(int *ip) {
int &ir = deref(ip);
}
Specific changes/additions:
* Template argument deduction from a call to a function template.
* Instantiation of a function template specializations (just the
declarations) from the template arguments deduced from a call.
* FunctionTemplateDecls are stored directly in declaration contexts
and found via name lookup (all forms), rather than finding the
FunctionDecl and then realizing it is a template. This is
responsible for most of the churn, since some of the core
declaration matching and lookup code assumes that all functions are
FunctionDecls.
llvm-svn: 74213
printing logic to help customize the output. For now, we use this
rather than a special flag to suppress the "struct" when printing
"struct X" and to print the Boolean type as "bool" in C++ but "_Bool"
in C.
llvm-svn: 72590
specifier resulted in the creation of a new TagDecl node, which
happens either when the tag specifier was a definition or when the tag
specifier was the first declaration of that tag type. This information
has several uses, the first of which is implemented in this commit:
1) In C++, one is not allowed to define tag types within a type
specifier (e.g., static_cast<struct S { int x; } *>(0) is
ill-formed) or within the result or parameter types of a
function. We now diagnose this.
2) We can extend DeclGroups to contain information about any tags
that are declared/defined within the declaration specifiers of a
variable, e.g.,
struct Point { int x, y, z; } p;
This will help improve AST printing and template instantiation,
among other things.
3) For C99, we can keep track of whether a tag type is defined
within the type of a parameter, to properly cope with cases like,
e.g.,
int bar(struct T2 { int x; } y) {
struct T2 z;
}
We can also do similar things wherever there is a type specifier,
e.g., to keep track of where the definition of S occurs in this
legal C99 code:
(struct S { int x, y; } *)0
llvm-svn: 72555
template<typename T>
struct X {
struct Inner;
};
template struct X<int>::Inner;
This change is larger than it looks because it also fixes some
a problem with nested-name-specifiers and tags. We weren't requiring
the DeclContext associated with the scope specifier of a tag to be
complete. Therefore, when looking for something like "struct
X<int>::Inner", we weren't instantiating X<int>.
This, naturally, uncovered a problem with member pointers, where we
were requiring the left-hand side of a member pointer access
expression (e.g., x->*) to be a complete type. However, this is wrong:
the semantics of this expression does not require a complete type (EDG
agrees).
Stuart vouched for me. Blame him.
llvm-svn: 71756
- This is a WIP...
- This adds -march= handling to the driver, and fixes the defaulting
of -mcpu on Darwin (which was using the wrong test).
Instead of handling -m{sse, ...} in the driver, pass them to clang-cc as
-target-feature [+-]name
In clang-cc, communicate with the (clang) target to discover the legal
features of a target, and the features which are enabled based on
-mcpu. This is currently hardcoded just enough to not be a feature
regression, we need to get this information from the backend's
TableGen information somehow.
This is used to construct the full list of features which are being
used, which is in turn used to initialize the predefines.
llvm-svn: 71061
in C++, taking into account conversions to the "composite pointer
type" so that we can compare, e.g., a pointer to a derived class to a
pointer to a base class.
Also, upgrade the "comparing distinct pointer types" from a warning to
an error for C++, since this is clearly an error. Turns out that we
hadn't gone through and audited this code for C++, ever.
Fixes <rdar://problem/6816420>.
llvm-svn: 70829
Fariborz Jahanian