iff its substitution contains an unexpanded parameter pack. This has the effect
that we now reject declarations such as this (which we used to crash when
expanding):
template<typename T> using Int = int;
template<typename ...Ts> void f(Int<Ts> ...ints);
The standard is inconsistent on how this case should be treated.
llvm-svn: 148905
pointer to incomplete type from an ExtWarn to an error. We put the
ExtWarn in place as part of a workaround for Boost (PR6527), but it
(1) doesn't actually match a GCC extension and (2) has been fixed for
two years in Boost, and (3) causes us to emit code that fails badly at
run time, so it's a bad idea to keep it. Fixes PR11803.
llvm-svn: 148838
not integer constant expressions. In passing, fix the 'folding is an extension'
diagnostic to not claim we're accepting the code, since that's not true in
-pedantic-errors mode, and add this diagnostic to -Wgnu.
llvm-svn: 148209
scope, when no other indication is provided that the user intended to declare a
function rather than a variable.
Remove some false positives from the existing 'parentheses disambiguated as a
function' warning by suppressing it when the declaration is marked as 'typedef'
or 'extern'.
Add a new warning group -Wvexing-parse containing both of these warnings.
The new warning is enabled by default; despite a number of false positives (and
one bug) in clang's test-suite, I have only found genuine bugs with it when
running it over a significant quantity of real C++ code.
llvm-svn: 147599
members of class templates so that their values can be used in ICEs. This
required reverting r105465, to get such instantiated members to be included in
serialized ASTs.
llvm-svn: 147023
diagnostic message are compared. If either is a substring of the other, then
no error is given. This gives rise to an unexpected case:
// expect-error{{candidate function has different number of parameters}}
will match the following error messages from Clang:
candidate function has different number of parameters (expected 1 but has 2)
candidate function has different number of parameters
It will also match these other error messages:
candidate function
function has different number of parameters
number of parameters
This patch will change so that the verification string must be a substring of
the diagnostic message before accepting. Also, all the failing tests from this
change have been corrected. Some stats from this cleanup:
87 - removed extra spaces around verification strings
70 - wording updates to diagnostics
40 - extra leading or trailing characters (typos, unmatched parens or quotes)
35 - diagnostic level was included (error:, warning:, or note:)
18 - flag name put in the warning (-Wprotocol)
llvm-svn: 146619
Necessary to parse Microsoft ATL code.
Example:
int array[] = {
0,
__if_exists(CLASS::Type) {2, }
3
};
will declare an array of 2 or 3 elements depending on if CLASS::Type exists or not.
llvm-svn: 146447
methods) to bool. E.g.
void foo() {}
if (f) { ... // <- Warns here.
}
Only applies to non-weak functions, and does not apply if the function address
is taken explicitly with the addr-of operator.
llvm-svn: 145849
Basically typo correction will try to offer a correction instead of looking into type dependent base classes.
I found this problem while parsing Microsoft ATL code with clang.
llvm-svn: 145772
Basically we have to look into the parent *lexical* DeclContext for friend functions at class scope. That's because calling GetParent() return the namespace or file DeclContext.
This fixes all remaining cases of "Unqualified lookup into dependent bases of class templates" when parsing MFC code with clang.
llvm-svn: 145127
semantics and defaults as the corresponding g++ arguments. The historical g++
argument -ftemplate-depth-N is kept for compatibility, but modern g++ versions
no longer document that option.
Add -cc1 argument -fconstexpr-depth N to implement the corresponding
functionality.
The -ftemplate-depth=N part of this fixes PR9890.
llvm-svn: 145045
This is a little bit tricky because during default argument instantiation the CurContext points to a CXXMethodDecl but we can't use the keyword this or have an implicit member call generated.
This fixes 2 errors when parsing MFC code with clang.
llvm-svn: 144881
which they do. This avoids all of the default argument promotions that
we (1) don't want, and (2) undo during that custom type checking, and
makes sure that we don't run into trouble during template
instantiation. Fixes PR11320.
llvm-svn: 144110
the injected-class-name of a class (or class template) to the
declaration that results from substituting the given template
arguments. Previously, we would actually perform a substitution into
the injected-class-name type and then retrieve the resulting
declaration. However, in certain, rare circumstances involving
deeply-nested member templates, we would get the wrong substitution
arguments.
This new approach just matches up the declaration with a declaration
that's part of the current context (or one of its parents), which will
either be an instantiation (during template instantiation) or the
declaration itself (during the definition of the template). This is
both more efficient (we're avoiding a substitution) and more correct
(we can't get the template arguments wrong in the member-template
case).
Fixes <rdar://problem/9676205>.
Reinstated, now that we have the fix in r143967.
llvm-svn: 143968
the injected-class-name of a class (or class template) to the
declaration that results from substituting the given template
arguments. Previously, we would actually perform a substitution into
the injected-class-name type and then retrieve the resulting
declaration. However, in certain, rare circumstances involving
deeply-nested member templates, we would get the wrong substitution
arguments.
This new approach just matches up the declaration with a declaration
that's part of the current context (or one of its parents), which will
either be an instantiation (during template instantiation) or the
declaration itself (during the definition of the template). This is
both more efficient (we're avoiding a substitution) and more correct
(we can't get the template arguments wrong in the member-template
case).
Fixes <rdar://problem/9676205>.
llvm-svn: 143551
does not match any declaration in the class (or class template), be
sure to mark it as invalid. Fixes PR10924 / <rdar://problem/10119422>.
llvm-svn: 143504
that it retains source location information for the type. Aside from
general goodness (being able to walk the types described in that
information), we now have a proper representation for dependent
delegating constructors. Fixes PR10457 (for real).
llvm-svn: 143410
Microsoft __if_exists/__if_not_exists statement. Also note that we
weren't traversing DeclarationNameInfo *at all* within the
RecursiveASTVisitor, which would be rather fatal for variadic
templates.
llvm-svn: 142906
statements. As noted in the documentation for the AST node, the
semantics of __if_exists/__if_not_exists are somewhat different from
the way Visual C++ implements them, because our parsed-template
representation can't accommodate VC++ semantics without serious
contortions. Hopefully this implementation is "good enough".
llvm-svn: 142901
analysis to separate dependent names from non-dependent names. For
dependent names, we'll behave differently from Visual C++:
- For __if_exists/__if_not_exists at class scope, we'll just warn
and then ignore them.
- For __if_exists/__if_not_exists in statements, we'll treat the
inner statement as a compound statement, which we only instantiate
in templates where the dependent name (after instantiation)
exists. This behavior is different from VC++, but it's as close as
we can get without encroaching ridiculousness.
The latter part (dependent statements) is not yet implemented.
llvm-svn: 142864
be sure to consider all of the possible lookup results. We were
assert()'ing (but behaving correctly) for unresolved values. Fixes
PR11134 / <rdar://problem/10290422>.
llvm-svn: 142652
actually just has an extraneous 'template<>' header, strip off the
'template<>' header and treat it as a normal friend tag. Fixes PR10660
/ <rdar://problem/9958322>.
llvm-svn: 142587
the right namespace in C++11 mode. Teach the code to prefer the 'must be in
precisely this namespace' diagnostic whenever that's true, and fix a defect
which resulted in the -Wc++11-compat warning in C++98 mode sometimes being
omitted.
llvm-svn: 142329
part of template argument deduction is ill-formed, we mark it as
invalid and treat it as a deduction failure. If we happen to find that
specialization again, treat it as a deduction failure rather than
silently building a call to the declaration.
Fixes PR11117, a marvelous bug where deduction failed after creating
an invalid specialization, causing overload resolution to pick a
different candidate. Then we performed a similar overload resolution
later, and happily picked the invalid specialization to
call... resulting in a silent link failure.
llvm-svn: 141809
We'd also like for "C++11" or "c++11" to be used for the warning
groups, but without removing the old warning flags. Patches welcome;
I've run out of time to work on this today.
llvm-svn: 141801
and DefaultFunctionArrayLvalueConversion. To prevent
significant regression for should-this-be-a-call fixits,
and to repair some such regression from the introduction of
bound member placeholders, make those placeholder checks
try to build calls appropriately. Harden the build-a-call
logic while we're at it.
llvm-svn: 141738
We had an extension which allowed const static class members of floating-point type to have in-class initializers, 'as a C++0x extension'. However, C++0x does not allow this. The extension has been kept, and extended to all literal types in C++0x mode (with a fixit to add the 'constexpr' specifier).
llvm-svn: 140801
Much to everyone's surprise, the default constructor for TypeResult produces
an instance with Invalid == false. This seems like a decision we may want to
revisit.
llvm-svn: 138601
even when overloaded and user-defined. These operators are both more
valuable to warn on (due to likely typos) and extremely unlikely to be
reasonable for use to trigger side-effects.
llvm-svn: 137823
-Wunused was a mistake. It resulted in duplicate warnings and lots of
other hacks. Instead, this should be a special sub-category to
-Wunused-value, much like -Wunused-result is.
Moved to -Wunused-comparison, moved the implementation to piggy back on
the -Wunused-value implementation instead of rolling its own, different
mechanism for catching all of the "interesting" statements.
I like the unused-value mechanism for this better, but its currently
missing several top-level statements. For now, I've FIXME-ed out those
test cases. I'll enhance the generic infrastructure to catch these
statements in a subsequent patch.
This patch also removes the cast-to-void fixit hint. This hint isn't
available on any of the other -Wunused-value diagnostics, and if we want
it to be, we should add it generically rather than in one specific case.
llvm-svn: 137822
Example:
template <class T>
class A {
public:
template <class U> void f(U p) { }
template <> void f(int p) { } // <== class scope specialization
};
This extension is necessary to parse MSVC standard C++ headers, MFC and ATL code.
BTW, with this feature in, clang can parse (-fsyntax-only) all the MSVC 2010 standard header files without any error.
llvm-svn: 137573
class C {
public:
static void foo2() { }
};
template <class T> class A {
public:
typedef C D;
};
template <class T> class B : public A<T> {
public:
void foo() { D::foo2(); }
};
Note that this won't work if the NestedNameSpecifier refers to a type.
This fixes 1 error when parsing the MSVC 2010 standard headers file with clang.
llvm-svn: 136203
that allocates an array of objects with a non-trivial destructor, be
sure to mark the destructor is "used". Fixes PR10480 /
<rdar://problem/9834317>.
llvm-svn: 136081
which is required given the current setup for template
argument deduction substitution validation, and add a test
case to make sure we don't break it in the future.
llvm-svn: 135262
template<typename T> struct S { } f() { return 0; }
This case now produces a missing ';' diagnostic, since that seems like a much more likely error than an attempt to declare a function or variable in addition to the class template.
Treat this
llvm-svn: 135195
were just punting on template argument deduction for a number of type
nodes. Most of them, obviously, didn't matter.
As a consequence of this, make extended vector types (via the
ext_vector_type attribute) actually work properly for several
important cases:
- If the attribute appears in a type-id (i.e, not attached to a
typedef), actually build a proper vector type
- Build ExtVectorType whenever the size is constant; previously, we
were building DependentSizedExtVectorType when the size was constant
but the type was dependent, which makes no sense at all.
- Teach template argument deduction to handle
ExtVectorType/DependentSizedExtVectorType.
llvm-svn: 133060
before the template parameters have acquired a proper context (e.g.,
because the enclosing context has yet to be built), provide empty
parameter lists for all outer template parameter scopes to inhibit any
substitution for those template parameters. Fixes PR9643 /
<rdar://problem/9251019>.
llvm-svn: 133055
- Move a test from test/SemaTemplate/instantiate-expr-3.cpp, it did not belong there
- Incomplete and abstract types are considered hard errors
llvm-svn: 132979
namespace set algorithm (re-)introduced. We may not have seen the 'std'
namespace, but we should still suggested associated namespaces. Easy
fix, but a bit annoying to test.
llvm-svn: 132744
Sema::RequireCompleteExprType() a bit more, setting the point of
instantiation if needed, and skipping explicit specializations entirely.
llvm-svn: 132547
of incomplete array type, attempt to complete the array type. This was
made much easier by Chandler's addition of RequireCompleteExprType(),
which I've tweaked (slightly) to improve the consistency of the
DeclRefExpr. Fixes PR7985.
llvm-svn: 132530
within class templates when they are necessary to complete the type of
the member. The canonical example is code like:
template <typename T> struct S {
static const int arr[];
static const int x;
static int f();
};
template <typename T> const int S<T>::arr[] = { 1, 2, 3 };
template <typename T> const int S<T>::x = sizeof(arr) / sizeof(arr[0]);
template <typename T> int S<T>::f() { return x; }
int x = S<int>::f();
We need to instantiate S<T>::arr's definition to pick up its initializer
and complete the array type. This involves new code to specially handle
completing the type of an expression where the type alone is
insufficient. It also requires *updating* the expression with the newly
completed type. Fortunately, all the other infrastructure is already in
Clang to do the instantiation, do the completion, and prune out the
unused bits of code that result from this instantiation.
This addresses the initial bug in PR10001, and will be a step to
fleshing out other cases where we need to work harder to complete an
expression's type. Who knew we still had missing C++03 "features"?
llvm-svn: 132172
also consider whether any of the parameter types (as written, prior to
decay) are dependent. Fixes PR9880 and <rdar://problem/9408413>.
llvm-svn: 131099
parameters on the floor in certain cases:
class X {
template <typename T> friend typename A<T>::Foo;
};
This was parsed as a *non* template friend declaration some how, and
received an ExtWarn. Fixing the parser to actually provide the template
parameters to the freestanding declaration parse triggers the code which
specifically looks for such constructs and hard errors on them.
Along the way, this prevents us from trying to instantiate constructs
like the above inside of a outer template. This is important as loosing
the template parameters means we don't have a well formed declaration
and template instantiation will be unable to rebuild the AST. That fixes
a crash in the GCC test suite.
llvm-svn: 130772
partial ordering of function templates, use a simple superset
relationship rather than the convertibility-implying
isMoreQualifiedThan/compatibilyIncludes relationship. Fixes partial
ordering between references and address-space-qualified references.
llvm-svn: 130612
in the classification of template names and using declarations. We now
properly typo-correct the leading identifiers in statements to types,
templates, values, etc. As an added bonus, this reduces the number of
lookups required for disambiguation.
llvm-svn: 130288
dealing with address-space- and GC-qualified pointers. Previously,
these qualifiers were being treated just like cvr-qualifiers (in some
cases) or were completely ignored, leading to uneven behavior. For
example, const_cast would allow conversion between pointers to
different address spaces.
The new semantics are fairly simple: reinterpret_cast can be used to
explicitly cast between pointers to different address spaces
(including adding/removing addresss spaces), while
static_cast/dynamic_cast/const_cast do not tolerate any changes in the
address space. C-style casts can add/remove/change address spaces
through the reinterpret_cast mechanism. Other non-CVR qualifiers
(e.g., Objective-C GC qualifiers) work similarly.
As part of this change, I tweaked the "casts away constness"
diagnostic to use the term "casts away qualifiers". The term
"constness" actually comes from the C++ standard, despite the fact
that removing "volatile" also falls under that category. In Clang, we
also have restrict, address spaces, ObjC GC attributes, etc., so the
more general "qualifiers" is clearer.
llvm-svn: 129583
The idea is that you can create a VarDecl with an unknown type, or a
FunctionDecl with an unknown return type, and it will still be valid to
access that object as long as you explicitly cast it at every use. I'm
still going back and forth about how I want to test this effectively, but
I wanted to go ahead and provide a skeletal implementation for the LLDB
folks' benefit and because it also improves some diagnostic goodness for
placeholder expressions.
llvm-svn: 129065
overload, so that we actually do the resolution for full expressions
and emit more consistent, useful diagnostics. Also fixes an IRGen
crasher, where Sema wouldn't diagnose a resolvable bound member
function template-id used in a full-expression (<rdar://problem/9108698>).
llvm-svn: 127747
to set the source-location information for the template arguments to
the *transformed* source-location information, not the original
source-location information. Fixes <rdar://problem/8986308> (a libc++
SFINAE issue) and the Boost.Polygon failure.
llvm-svn: 127150
declaration because of interesting ordering dependencies while
instantiating a class template or member class thereof. Complain,
rather than asserting (+Asserts) or silently rejecting the code
(-Asserts).
Fixes the crash-on-invalid in PR8965.
llvm-svn: 127129
template name as the result of substitution. The qualifier is handled
separately by the tree transformer, so we would end up in an
inconsistent state.
This is actually the last bit of PR9016, and possibly also fixes
PR8965. It takes Boost.Icl from "epic fail" down to a single failure.
llvm-svn: 127108
template (not a specialization!), use the "injected" function template
arguments, which correspond to the template parameters of the function
template. This is required when substituting into the default template
parameters of template template parameters within a function template.
Fixes PR9016.
llvm-svn: 127092
transform the type that replaces the template type parameter. In the
vast majority of cases, there's nothing to do, because most template
type parameters are replaced with something non-dependent that doesn't
need further transformation. However, when we're dealing with the
default template arguments of template template parameters, we might
end up replacing a template parameter (of the template template
parameter) with a template parameter of the enclosing template.
This addresses part of PR9016, but not within function
templates. That's a separate issue.
llvm-svn: 127091
to cope with non-type templates by providing appropriate
errors. Previously, we would either assert, crash, or silently build a
dependent type when we shouldn't. Fixes PR9226.
llvm-svn: 127037
DependentTemplateSpecializationType during tree transformation, retain
the NestedNameSpecifierLoc as it was used to translate the template
name, rather than reconstructing it from the template name.
Fixes PR9401.
llvm-svn: 127015
DeclContext once we've created it. This mirrors what we do for
function parameters, where the parameters start out with
translation-unit context and then are adopted by the appropriate
DeclContext when it is created. Also give template parameters public
access and make sure that they don't show up for the purposes of name
lookup.
Fixes PR9400, a regression introduced by r126920, which implemented
substitution of default template arguments provided in template
template parameters (C++ core issue 150).
How on earth could the DeclContext of a template parameter affect the
handling of default template arguments?
I'm so glad you asked! The link is
Sema::getTemplateInstantiationArgs(), which determines the outer
template argument lists that correspond to a given declaration. When
we're instantiating a default template argument for a template
template parameter within the body of a template definition (not it's
instantiation, per core issue 150), we weren't getting any outer
template arguments because the context of the template template
parameter was the translation unit. Now that the context of the
template template parameter is its owning template, we get the
template arguments from the injected-class-name of the owning
template, so substitution works as it should.
llvm-svn: 127004
parameter, save the instantiated default template arguments along with
the explicitly-specified template argument list. That way, we prefer
the default template template arguments corresponding to the template
template parameter rather than those of its template template argument.
This addresses the likely direction of C++ core issue 150, and fixes
PR9353/<rdar://problem/9069136>, bringing us closer to the behavior of
EDG and GCC.
llvm-svn: 126920
template specialization types. This also required some parser tweaks,
since we were losing track of the nested-name-specifier's source
location information in several places in the parser. Other notable
changes this required:
- Sema::ActOnTagTemplateIdType now type-checks and forms the
appropriate type nodes (+ source-location information) for an
elaborated-type-specifier ending in a template-id. Previously, we
used a combination of ActOnTemplateIdType and
ActOnTagTemplateIdType that resulted in an ElaboratedType wrapped
around a DependentTemplateSpecializationType, which duplicated the
keyword ("class", "struct", etc.) and nested-name-specifier
storage.
- Sema::ActOnTemplateIdType now gets a nested-name-specifier, which
it places into the returned type-source location information.
- Sema::ActOnDependentTag now creates types with source-location
information.
llvm-svn: 126808
a dependent template name rather than (indirectly and incorrectly)
trying to determine whether we can compute a context for the
nested-name-specifier. Fixes a GCC testsuite regression,
<rdar://problem/9068589>.
llvm-svn: 126749
current instantiation, even though we have a RecordDecl describing
them. Fixes PR9255.
Amusingly, I've had this patch sitting around for a month or two
because it was "obviously" wrong, but hadn't gotten around to writing
a test case to submit the fix :)
llvm-svn: 126038
or union, place the qualifier on the outermost member reference
expression, which actually contains the entity name.
Fixes PR9188/<rdar://problem/8990184>.
llvm-svn: 125822
enumeration type, we were generating an integer literal implicitly
casted to the appropriate enumeration type. However, later checks on
that expression would strip the implicit cast.
This commit tweaks the lame hack, by creating an explicit cast instead
of an implicit cast. The right answer is to introduce a
SubstNonTypeTemplateParmExpr expression that acts like the substituted
result. I'll investigate that soon.
llvm-svn: 125818
parameter type to see what's behind it, so that we don't end up
printing silly things like "float const *" when "const float *" would
make more sense. Also, replace the pile of "isa" tests with a simple
switch enumerating all of the cases, making a few more obvious cases
use prefix qualifiers.
llvm-svn: 125729
id-expression, e.g.,
CurrentClass<T>::member
Previously, if CurrentClass<T> was dependent and not complete, we
would treat it as a dependent-scoped declaration reference expression,
even if CurrentClass<T> referred to the current instantiation.
Fixes PR8966 and improves type checking of templates.
llvm-svn: 124867
parameters into parameter types, so that substitution of
explicitly-specified function template arguments uses the same
path. This enables the use of explicitly-specified function template
arguments with variadic templates.
llvm-svn: 122986
declaration name of the array when present. This ensures that
a poor-man's C++03 static_assert will include the user error message
often embedded in the name.
Update all the tests to reflect the new wording, and add a test for the
name behavior.
llvm-svn: 122802
We should not substitute template types if the template has a dependent
context because the template argument stack is not yet fully formed.
Instead, defer substitution until the template has a non-dependent
context (i.e. instantiation of an outer template).
llvm-svn: 121491
declaration that is a value in ill-formed code. Instead of crashing,
treat this as a dependent typename specifier and suggest that the
using add "typename" into the using declaration. Fixes <rdar://problem/8740998>.
llvm-svn: 121322
build one when either of the operands calls itself type-dependent;
previously we were building when one of the operand types was dependent,
which is not always the same thing and which can lead to unfortunate
inconsistencies later. Fixes PR8739.
llvm-svn: 120990
not actually frequently used, because ImpCastExprToType only creates a node
if the types differ. So explicitly create an ICE in the lvalue-to-rvalue
conversion code in DefaultFunctionArrayLvalueConversion() as well as several
other new places, and consistently deal with the consequences throughout the
compiler.
In addition, introduce a new cast kind for loading an ObjCProperty l-value,
and make sure we emit those nodes whenever an ObjCProperty l-value appears
that's not on the LHS of an assignment operator.
This breaks a couple of rewriter tests, which I've x-failed until future
development occurs on the rewriter.
Ted Kremenek kindly contributed the analyzer workarounds in this patch.
llvm-svn: 120890
instantiations, GCC also supports "inline" and "static" explicit
template instantiations. Parse and warn about such constructs, but
don't implement the semantics of either "inline" or "static". They
don't seem to be widely used.
llvm-svn: 120599
(while computing user conversion sequences), make sure that a result
of class type is a complete class type. Had we gone through
ActOnCallExpr, this would have happened when we built the CallExpr.
Fixes PR8425.
llvm-svn: 119005
parameters to the Transform*Type functions and instead call out
the specific cases where an object type and the unqualified lookup
results are important. Fixes an assert and failed compile on
a testcase from PR7248.
llvm-svn: 118887
in the order they occur within the class template, delaying
out-of-line member template partial specializations until after the
class has been fully instantiated. This fixes a regression introduced
by r118454 (itself a fix for PR8001).
llvm-svn: 118704
constructor template will not be used to copy a class object to a
value of its own type. We were eliminating all constructor templates
whose specializations look like a copy constructor, which eliminated
important candidates. Fixes PR8182.
llvm-svn: 118418
abstractions (e.g., TemplateArgumentListBuilder) that were designed to
support variadic templates. Only a few remnants of variadic templates
remain, in the parser (parsing template type parameter packs), AST
(template type parameter pack bits and TemplateArgument::Pack), and
Sema; these are expected to be used in a future implementation of
variadic templates.
But don't get too excited about that happening now.
llvm-svn: 118385
of its parent context, be sure to update the parent-context pointer
after instantiation. Fixes two anonymous-union instantiation issues in
<rdar://problem/8635664>.
llvm-svn: 118313
or dependent specializations, rip apart the dependent name/dependent
specialization to recanonicalize its pieces, because
nested-name-specifiers store "dependent-type::identifier" differently
than types do. Fixes PR7419.
llvm-svn: 118211
ambiguous name where none of the declarations found are actually
templates. In this case, make sure we clear out the ambiguous-path
data when recomputing the lookup result kind. Fixes PR8439.
llvm-svn: 117112
within a default argument), recurse into default arguments. Fixes
PR8401, a regression I introduced in r113700 while refactoring our
handling of "used" declarations in default arguments.
llvm-svn: 116817
that the class type into which the pointer points be complete, even
though the standard requires it. GCC/EDG do not require a complete
type here, so we're calling this a problem with the standard. Fixes
PR8328.
llvm-svn: 116429
Richard Smith