just valid C++11 =)
Original commit message:
PR18427: Use an appropriately-aligned buffer in APValue, to avoid a crash on
SPARC, where uint64_t apparently requires higher alignment than void*.
llvm-svn: 198903
The MS abi lays out *all* non-virtual bases with leading vfptrs before
laying out non-virutal bases without vfptrs. This guarantees that the
primary base is laid out first. r198818 fixed RecordLayoutBuilder to
produce compatiable layouts. This patch fixes CGRecordLayoutBuilder to
be able to consume those layouts and produce meaningful output without
tripping any asserts about assumed incoming layout.
A test case is included that shows CGRecordLayoutBuilder in fact
produces output in the compatiable order.
llvm-svn: 198900
This patch refactors microsoft record layout to be more "natural". The
most dominant change is that vbptrs and vfptrs are injected after the
fact. This simplifies the implementation and the math for the offest
for the first base/field after the vbptr.
llvm-svn: 198818
encodes the canonical rules for LLVM's style. I noticed this had drifted
quite a bit when cleaning up LLVM, so wanted to clean up Clang as well.
llvm-svn: 198686
enum-scoped.cpp:93:6: error: enumeration redeclared with different underlying type 'short' (was 'int')
enum Redeclare6 : short;
^
enum-scoped.cpp:92:6: note: previous declaration is here
enum Redeclare6 : int;
^ ~~~
The redeclaration source range is still missing but this is a step forward,
potentially edging towards a FixIt.
llvm-svn: 198601
This has the dual effect of (1) enabling more dead-stripping in release builds
and (2) ensuring that debug helper functions aren't stripped away in debug
builds, as they're intended to be called from the debugger.
Note that the attribute is applied to definitions rather than declarations in
headers going forward because it's now conditional on NDEBUG:
/// \brief Mark debug helper function definitions like dump() that should not be
/// stripped from debug builds.
Requires corresponding macro added in LLVM r198456.
llvm-svn: 198489
Summary:
This makes us more compatible with MSVC 2012+ and fixes PR17748 where we
would give two tables the same name.
Rather than doing a fresh depth-first traversal of the inheritance graph
for every record's vbtables, now we memoize vbtable paths for each
record. By doing memoization, we end up considering virtual bases of
subobjects that come later in the depth-first traversal. Where
previously we would have ignored a virtual base that we'd already seen,
we now consider it for name mangling purposes without emitting a
duplicate vbtable for it.
Reviewers: majnemer
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D2509
llvm-svn: 198462
- Remove the additions to ObjCMethodDecl & ObjCIVarDecl that were getting de/serialized and consolidate
all functionality for the checking for this warning in Sema::DiagnoseUnusedBackingIvarInAccessor
- Don't check immediately after the method body is finished, check when the @implementation is finished.
This is so we can see if the ivar was referenced by any other method, even if the method was defined after the accessor.
- Don't silence the warning if any method is called from the accessor silence it if the accessor delegates to another method via self.
rdar://15727325
llvm-svn: 198432
Summary:
No functionality change.
This code should live here long-term because we should be able to use it
to compute correct vftable names.
It turns out that the most natural way to implement the naming algorithm
is to use a caching layer similar to what we already have for virtual
table info in VTableContext. Subsequent changes will take advantage of
this to fix PR17748, where we have a vbtable name collision.
Reviewers: majnemer
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D2499
llvm-svn: 198380
Remove UnaryTypeTraitExpr and switch all remaining type trait related handling
over to TypeTraitExpr.
The UTT/BTT/TT enum prefix and evaluation code is retained pending further
cleanup.
This is part of the ongoing work to unify type traits following the removal of
BinaryTypeTraitExpr in r197273.
llvm-svn: 198271
important for thread safety attributes, which contain expressions that were
not being visited, and were thus invisible to various tools. There are now
Visit*Attr methods that can be overridden for every attribute.
llvm-svn: 198224
Most importantly, this makes our vtable layout match MSVC's. Previously
we would emit a return adjusting thunk whenever the return types
differed, even if the adjustment would have been trivial.
MSVC does emit some trivial return adjusting thunks, but only if there
was already an overridden method that required a return adjustment.
llvm-svn: 198080
With pragma pack, the layout engine would produce vfptrs that were
packed width rather than pointer width. This patch addresses the issue
and adds a test case.
llvm-svn: 198059
Now CodeGenVTables has only one VTableContext object, which is either
Itanium or Microsoft.
Fixes a FIXME with no functionality change intended.
Ideally we could avoid the downcasts by pushing the things that
reference the Itanium vtable context into ItaniumCXXABI.cpp, but we're
not there yet.
llvm-svn: 197845
This was part of the cause for PR17655. We were generating thunks when
we shouldn't have. I suspect that if we tweak the test case for PR17655
to actually require thunks, we can reproduce the same crash.
llvm-svn: 197836
The alignment impact of the virtual bases apperas to be applied in
order, rather than up front. This patch adds the new behavior and
provides a test case.
llvm-svn: 197639
__builtin_va_list and friends have been showing up where they shouldn't for way
to long, making unwanted appearences in -ast-print, tooling and source level
visitors and even the hello world tutorial on the clang website.
This commit factors down the implicit typedef and record creation facilities to
ensure they're marked implicit.
Also fixes a unit test that was testing incorrect behaviour, and removes old
hacks in the DeclPrinter that tried to skip implicit declarations manually.
llvm-svn: 197336
There's nothing special about type traits accepting two arguments.
This commit eliminates BinaryTypeTraitExpr and switches all related handling
over to TypeTraitExpr.
Also fixes a CodeGen failure with variadic type traits appearing in a
non-constant expression.
The BTT/TT prefix and evaluation code is retained as-is for now but will soon
be further cleaned up.
This is part of the ongoing work to unify type traits.
llvm-svn: 197273
The old URL hasn't worked for quite some time. While we are here, also
change the link so that it will send us straight to the mangling portion
of the ABI doc.
llvm-svn: 197195
This reverts commit r197184.
Richard Smith brings up some good points, a proper implementation will
require us to mangle unnameable entities compatibly with MSVC.
llvm-svn: 197192
This refactor addresses bugzilla bug 18167 and simplifies the code at
the same time. Also I add a test case for the bug. Also I make a
non-functional change to the basic layout lit tests to make them more
reliable (using CHECK-NEXT instead of CHECK).
llvm-svn: 197183
Type trait parsing is all over the place at the moment with unary, binary and
n-ary C++11 type traits that were developed independently at different points
in clang's history.
There's no good reason to handle them separately -- there are three parsers,
three AST nodes and lots of duplicated handling code with slightly different
implementations and diags for each kind.
This commit unifies parsing of type traits and sets the stage for further
consolidation.
No change in behaviour other than more consistent error recovery.
llvm-svn: 197179
declarations that might lifetime-extend multiple temporaries. In passing, fix a
crasher (PR18217) if an initializer was dependent and exactly the wrong shape,
and remove a bogus function (Expr::findMaterializedTemporary) now its last use
is gone.
llvm-svn: 197103
After r196549 there is no need to separate FinalizeCXXLayout and
FinalizeLayout so they were merged and FinalizeCXXLayout was eliminated.
llvm-svn: 197083
With the introduction of explicit address space casts into LLVM, there's
a need to provide a new cast kind the front-end can create for C/OpenCL/CUDA
and code to produce address space casts from those kinds when appropriate.
Patch by Michele Scandale!
llvm-svn: 197036
Prior to this patch, the alignment imposed by virtual bases only
included direct virtual bases. This patch fixes it to look at all
virtual bases.
llvm-svn: 196997
That's a mouthful, and not necessarily the final name. This also
reflects a semantic change where this attribute is now on the
protocol itself instead of a class. This attribute will require
that a protocol, when adopted by a class, is explicitly implemented
by the class itself (instead of walking the super class chain).
Note that this attribute is not "done". This should be considered
a WIP.
llvm-svn: 196955
more than one such initializer in a union, make mem-initializers override
default initializers for other union members, handle anonymous unions with
anonymous struct members better. Fix a couple of semi-related bugs exposed by
the tests for same.
llvm-svn: 196892
In order to address latent bugs that were easier to expose in 64-bit
mode, we move the application of __declspec(align) to before the layout
of vbases rather than after.
llvm-svn: 196861
Testing has revealed that large integral constants (i.e. > INT64_MAX)
are always mangled as-if they are negative, even in places where it
would not make sense for them to be negative (like non-type template
parameters of type unsigned long long).
To address this, we change the way we model number mangling: always
mangle as-if our number is an int64_t. This should result in correct
results when we have large unsigned numbers.
N.B. Bizarrely, things that are 32-bit displacements like vbptr offsets
are mangled as-if they are unsigned 32-bit numbers. This is a pretty
egregious waste of space, it would be a 4x savings if we could mangle it
like a signed 32-bit number. Instead, we explicitly cast these
displacements to uint32_t and let the mangler proceed.
llvm-svn: 196771
While testing our ability to mangle large constants (PR18175), I
incidentally discovered that we did not properly mangle enums correctly.
Previously, we would append the width of the enum in bytes after the
type-tag differentiator.
This would mean "enum : short" would be mangled as 'W2' while "enum :
char" would be mangled as 'W1'. Upon testing this with several versions
of MSVC, I found that this did not match their behavior: they always use
'W4'.
N.B. Quick testing uncovered that undname allows different numbers to
follow the 'W' in the following way:
'W0' -> "enum char"
'W1' -> "enum unsigned char"
'W2' -> "enum short"
'W3' -> "enum unsigned short"
'W4' -> "enum"
'W5' -> "enum unsigned int"
'W6' -> "enum long"
'W7' -> "enum unsigned long"
However this scheme appears abandoned, I cannot get MSVC to trigger it.
Furthermore, it's incomplete: it doesn't handle "bool" or "long long".
llvm-svn: 196752
Clang outputs LLVM one top level decl at a time. This combined with the
visibility computation code looking for the newest NamespaceDecl would cause
it to produce different results for nested namespaces.
The two options for producing consistent results are
* Delay codegen of anything inside a namespace until the end of the file.
* Don't look for the newest NamespaceDecl.
This patch implements the second option.
This matches the gcc behavior too.
llvm-svn: 196712
MS-ABI adds padding before *every* vbase if the last field in a record
is a bit-field. This changes clangs behavior to match. I also fix some
windows-style line endings in the test file.
Differential Revision: http://llvm-reviews.chandlerc.com/D2277
llvm-svn: 196605
Adds padding between bases or virtual bases in an attempt to avoid
aliasing of zero-sized sub-objects. The approach used by the ABI adds
two more bits of state. Detailed comments are in the code. Test cases
included.
Differential Revision: http://llvm-reviews.chandlerc.com/D2258
llvm-svn: 196602
__declspec(align())
This patch implements required alignment in a way that makes
__declspec(align()) and #pragma pack play correctly together. In the
MS-ABI, __declspec(align()) is a hard rule and cannot be overridden by
#pragma pack. This cases each record to have two interesting alignments
"preferred alignment" (which matches Itanium's concept of alignment) and
"required alignment" which is an alignment that must never be violated,
even in the case of #pragma pack. This patch introduces the concept of
Required Alignment to the record builder and tracks/uses it
appropriately. Test cases are included.
Differential Revision: http://llvm-reviews.chandlerc.com/D2283
llvm-svn: 196549
don't assume that it inherits the designated initializers from the super class.
If the assumption was wrong because a new initializer was a designated one that was not marked as such,
we will emit misleading warnings for subclasses of the interface.
llvm-svn: 196476
Summary:
In general, this type node can be used to represent any type adjustment
that occurs implicitly without losing type sugar. The immediate use of
this is to adjust the calling conventions of member function pointer
types without breaking template instantiation.
Fixes PR17996.
Reviewers: rsmith
Differential Revision: http://llvm-reviews.chandlerc.com/D2332
llvm-svn: 196451
super another initializer and when the implementation does not delegate to
another initializer via a call on 'self'.
A secondary initializer is an initializer method not marked as a designated
initializer within a class that has at least one initializer marked as a
designated initializer.
llvm-svn: 196318
designated initializers of an interface.
If the interface declaration does not have methods marked as designated
initializers then the interface inherits the designated initializers of
its super class.
llvm-svn: 196315
It wasn't possible for an anonymous type to show up inside of function arguments.
However, decltype (which MSVC added support for in 2010) makes this
possible. Further, backrefs to these anonymous types can now be formed.
This fixes PR18022.
N.B. We do not, and very likely _will not_, support MSVC's bug where
subsequent typedefs of anonymous types leak into the linkage name; this
is a gross violation of the ABI. A warning should be introduced to
inform our users of this particular shortcoming.
llvm-svn: 195669
This is still an experimental attribute, but I wanted it in tree
for review. It may still get yanked.
This attribute can only be applied to a class @interface, not
a class extension or category. It does not change the type
system rules for Objective-C, but rather the implementation checking
for Objective-C classes that explicitly conform to a protocol.
During protocol conformance checking, clang recursively searches
up the class hierarchy for the set of methods that compose
a protocol. This attribute will cause the compiler to not consider
the methods contributed by a super class, its categories, and those
from its ancestor classes. Thus this attribute is used to force
subclasses to redeclare (and hopefully re-implement) methods if
they decide to explicitly conform to a protocol where some of those
methods may be provided by a super class.
This attribute intentionally leaves out properties, which are associated
with state. This attribute only considers methods (at least right now)
that are non-property accessors. These represent methods that "do something"
as dictated by the protocol. This may be further refined, and this
should be considered a WIP until documentation gets written or this
gets removed.
llvm-svn: 195533
This enables a micro-optimization in protocol conformance checking
to not examine the class hierarchy twice per method.
As part of this change, remove the default arguments from lookupInstanceMethod()
and lookupClassMethod(). It was becoming very redundant. For clients
needing the default arguments, have them use the full API instead of
these convenience methods.
llvm-svn: 195532
can't accidentally be allocated the wrong way (missing prefix data for decls
from AST files, for instance) and simplifies the CreateDeserialized functions a
little. An extra DeclContext* parameter to the not-from-AST-file operator new
allows us to ensure that we don't accidentally call the wrong one when
deserializing (when we don't have a DeclContext), allows some extra checks, and
prepares for some planned modules-related changes to Decl allocation.
No functionality change intended.
llvm-svn: 195426
After implementing this patch, a few concerns about the language
feature itself emerged in my head that I had previously not considered.
I want to resolve those design concerns first before having
a half-designed language feature in the tree.
llvm-svn: 195328
The idea is to allow a class to stipulate that its methods (and those
of its parents) cannot be used for protocol conformance in a subclass.
A subclass is then explicitly required to re-implement those methods
of they are present in the class marked with this attribute.
Currently the attribute can only be applied to an @interface, and
not a category or class extension. This is by design. Unlike
protocol conformance, where a category can add explicit conformance
of a protocol to class, this anti-conformance really needs to be
observed uniformly by all clients of the class. That's because
the absence of the attribute implies more permissive checking of
protocol conformance.
This unfortunately required changing method lookup in ObjCInterfaceDecl
to take an optional protocol parameter. This should not slow down
method lookup in most cases, and is just used for protocol conformance.
llvm-svn: 195323
Summary:
RTTI is not yet implemented for the Microsoft C++ ABI and isn't expected
soon. We could easily add the mangling, but the error is what prevents
us from silently miscompiling code that expects RTTI.
Instead, add a new mangleTypeName entry point that simply forwards to
mangleName or mangleType to produce a string that isn't part of the ABI.
Itanium can continue to use RTTI names to avoid unecessary test
breakage.
This also seems like the right design. The fact that TBAA names happen
to be RTTI names is now an implementation detail of the mangler, rather
than part of TBAA.
Differential Revision: http://llvm-reviews.chandlerc.com/D2153
llvm-svn: 195168
Microsoft adds an extra byte of padding before laying out zero sized
non-virtual bases if the non-virtual base before it contains a vbptr.
This patch adds the same behavior to clang.
Differential Revision: http://llvm-reviews.chandlerc.com/D2106
llvm-svn: 195158
Instead of storing the vtable offset directly in the function pointer and
doing a branch to check for virtualness at each call site, the MS ABI
generates a thunk for calling the function at a specific vtable offset,
and puts that in the function pointer.
This patch adds support for emitting such thunks. However, it doesn't support
pointers to virtual member functions that are variadic, have an incomplete
aggregate return type or parameter, or are overriding a function in a virtual
base class.
Differential Revision: http://llvm-reviews.chandlerc.com/D2104
llvm-svn: 194827
where we didn't. Extend our constant evaluation for __builtin_strlen to handle
any constant array of chars, not just string literals, to match.
llvm-svn: 194762
template, that member has a dependent type (even if we can see the definition
of the member of the primary template), because the array size could change in
a member specialization.
Patch by Karthik Bhat!
llvm-svn: 194740
bit fields of zero size. Warnings are generated in C++ mode and if
only such type is defined inside extern "C" block.
The patch fixed PR5065.
Differential Revision: http://llvm-reviews.chandlerc.com/D2151
llvm-svn: 194653
The xcore llvm backend does not handle 8 byte alignment viz:
"%BadAlignment = alloca i64, align 8"
So getPreferredTypeAlign() must never overalign.
llvm-svn: 194462
This makes it consistent with -fdump-record-layouts, which was moved to
outs() in r186219. My reasoning for going with stdout is that when one
of these options is present, the layouts are really a program output,
and shouldn't be interleaved with diagnostics, which are on stderr.
Reviewers: timurrrr
Differential Revision: http://llvm-reviews.chandlerc.com/D2127
llvm-svn: 194279
Summary:
Similar to __FUNCTION__, MSVC exposes the name of the enclosing mangled
function name via __FUNCDNAME__. This implementation is very naive and
unoptimized, it is expected that __FUNCDNAME__ would be used rarely in
practice.
Reviewers: rnk, rsmith, thakis
CC: cfe-commits, silvas
Differential Revision: http://llvm-reviews.chandlerc.com/D2109
llvm-svn: 194181
These functions can generally be applied to multiple kinds of AST node,
so it makes sense to add them to DynTypedNode.
Differential Revision: http://llvm-reviews.chandlerc.com/D2096
llvm-svn: 194113
bit more robust against future changes. This includes a slight diagnostic
improvement: if we know we're only trying to form a constant expression, take
the first diagnostic which shows the expression is not a constant expression,
rather than preferring the first one which makes the expression unfoldable.
llvm-svn: 194098
deallocation function (and the corresponding unsized deallocation function has
been declared), emit a weak discardable definition of the function that
forwards to the corresponding unsized deallocation.
This allows a C++ standard library implementation to provide both a sized and
an unsized deallocation function, where the unsized one does not just call the
sized one, for instance by putting both in the same object file within an
archive.
llvm-svn: 194055
would be deleted are still declared, but are ignored by overload resolution.
Also, don't delete such members if a subobject has no corresponding move
operation and a non-trivial copy. This causes us to implicitly declare move
operations in more cases, but risks move-assigning virtual bases multiple
times in some circumstances (a warning for that is to follow).
llvm-svn: 193969
If the sole distinction between two declarations is that one has a
__restrict qualifier then we should not consider it to be an overload.
Instead, we will consider it as an incompatible redeclaration which is
similar to how MSVC, ICC and GCC would handle it.
This fixes PR17786.
N.B. We must not mangle in __restrict into method qualifiers becase we
don't allow overloading between such declarations anymore. To do
otherwise would be a violation of the Itanium ABI.
llvm-svn: 193964
Differential Revision: http://llvm-reviews.chandlerc.com/D2090
Clang was "improperly" over-aligning arrays with sizes are not a multiple of
their alignment.
This behavior was removed in microsoft 32 bit mode.
In addition, after examination of ASTContext::getTypeInfoImpl, a redundant code block in
MicrosoftRecordLayoutBuilder::getAdjustedFieldInfo was deleted.
llvm-svn: 193898
Differential Revision: http://llvm-reviews.chandlerc.com/D2082
Adds a lang_c LinkageSpecDecl to lazily generated builtins. This enforces correct
behavior for builtins in a variety of cases without special treatment elsewhere within
the compiler (special treatment is removed by the patch). It also allows for C++
overloads of builtin functions, which Microsoft uses in their headers e.g.
_InterlockedExchangeAdd is an extern C builtin for the long type but an inline wrapper
for int type.
llvm-svn: 193896
it. Also removes all of the microsoft C++ ABI related code from the
itanium layout builder.
Differential Revision: http://llvm-reviews.chandlerc.com/D2003
llvm-svn: 193290
The Itanium mangler couldn't cope with mangling an IndirectFieldDecl.
Instead, mangle the field the IndirectFieldDecl refers to.
Further, give IndirectFieldDecl no linkage just like FieldDecl.
N.B. Decl.cpp:getLVForNamespaceScopeDecl tried to calculate linkage for
data members of anonymous structs/unions. However, this seems
impossible so turn it into an assertion.
llvm-svn: 193269
A prior commit of this patch was reverted because it was within the blamelist's purview of a failing test. The failure of that test has been addressed here: http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20131021/091546.html. Therefore I am recommitting this patch (all tests pass on windows, except for the usual modules & index suspects that never pass on my box).
Some background: Both Doug and Richard had asked me in Chicago to remove the circular reference in CXXRecordDecl to LambdaExpr by factoring out and storing the needed information from LambdaExpr directly into CXXRecordDecl.
In addition, I have added an IsGenericLambda flag - this makes life a little easier when we implement capturing, and are Sema-analyzing the body of a lambda (and the calloperator hasn't been wired to the closure class yet). Any inner lambdas can have potential captures that could require walking up the scope chain and checking if any generic lambdas are capture-ready. This 'bit' makes some of that checking easier.
No change in functionality.
This patch was approved by Doug with minor modifications (comments were cleaned up, and all data members were converted from bool/enum to unsigned, as requested):
http://llvm-reviews.chandlerc.com/D1856
Thanks!
llvm-svn: 193246
They were causing CodeGenCXX/mangle-exprs.cpp to fail.
Revert "Remove the circular reference to LambdaExpr in CXXRecordDecl."
Revert "Again: Teach TreeTransform and family how to transform generic lambdas nested within templates and themselves."
llvm-svn: 193226
Both Doug and Richard had asked me to remove the circular reference in CXXRecordDecl to LambdaExpr by factoring out and storing the needed information from LambdaExpr directly into CXXRecordDecl.
No change in functionality.
In addition, I have added an IsGenericLambda flag - this makes life a little easier when we implement capturing, and are Sema-analyzing the body of a lambda (and the calloperator hasn't been wired to the closure class yet). Any inner lambdas can have potential captures that could require walking up the scope chain and checking if any generic lambdas are capture-ready. This 'bit' makes some of that checking easier.
This patch was approved by Doug with minor modifications (comments were cleaned up, and all data members were converted from bool/enum to unsigned, as requested):
http://llvm-reviews.chandlerc.com/D1856
Thanks!
llvm-svn: 193223
This fixes pr17639.
Before this patch clang would consider
void foo(void) __attribute((alias("__foo")));
a declaration. It now correctly handles it as a definition.
Initial patch by Alp Toker. I added support for variables.
llvm-svn: 193200
* NamedDecl and CXXMethodDecl were missing getMostRecentDecl.
* The const version can just forward to the non const.
* getMostRecentDecl can use cast instead of cast_or_null.
This then removes some casts from the callers.
llvm-svn: 193039
Summary: Some MS headers use these features.
Reviewers: rnk, rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1948
llvm-svn: 192936
This removes the dependency on the llvm mangler doing it for us. In isolation,
the benefit is that the testing of what mangling is applied is all in one place:
(C, C++) X (Itanium, Microsoft) are all handled by clang.
This also gives me hope that in the future the llvm mangler (and llvm-ar) will
not depend on TargetMachine.
llvm-svn: 192762
If unqualified id lookup fails while parsing a class template with a
dependent base, clang with -fms-compatibility will pretend the user
prefixed the name with 'this->' in order to delay the lookup. However,
if there was a unary ampersand, Sema::ActOnDependentIdExpression() will
create a DependentDeclRefExpr, which is not what we wanted at all. Fix
this by building the CXXDependentScopeMemberExpr directly instead.
In order to be fully MSVC compatible, we would have to defer all
attempts at name lookup to instantiation time. However, until we have
real problems with system headers that can't be parsed, we'll put off
implementing that.
Fixes PR16014.
Reviewers: rsmith
Differential Revision: http://llvm-reviews.chandlerc.com/D1892
llvm-svn: 192727
We have to reserve at least the width of a pointer for the vfptr. For
classes with small alignment, we weren't reserving enough space, and
were overlapping the first field with the vfptr.
llvm-svn: 192626
This patch fixes the distructor test when checking for vtordisp requirements in
microsoft record layout. A test case is also included.
Addresses:
http://llvm.org/bugs/show_bug.cgi?id=16406#c7
llvm-svn: 192616
ASTImporter when importing the following types:
typedef struct {
} A;
typedef struct {
A a;
} B;
Suppose we have imported B, but we did not at that
time need to complete it. Then later we want to
import A. The struct is anonymous, so the first
thing we want to do is make sure no other anonymous
struct already matches it. So we set up an
StructuralEquivalenceContext and compare B with A.
This happens at ASTImporter.cpp:2179.
Now, in this scenario, B is not complete. So we go
and import its fields, including a, which causes A
to be imported. The ASTImporter doesn’t yet have A
in its list of already-imported things, so we
import A.
After the StructuralEquivalenceContext is finished
determining that A and B are different, the
ASTImporter concludes that A must be imported
because no equivalent exists, so it imports a second
copy of A. Now we have two different structs
representing A. This is really bad news.
The patch allows the StructuralEquivalenceContext to
use the original version of B when making its
comparison, obviating the need for an import and
cutting this loop.
llvm-svn: 192324
In the test case one type is coming from a typedef with no default arg, the
other has the default arg. Taking the default arg from the typedef crashes, so
always use the real template paramter declaration. PR17510.
llvm-svn: 192202
As described by Richard in https://groups.google.com/a/isocpp.org/d/msg/std-discussion/S1kmj0wF5-g/fb6agEYoL2IJ
we should allow:
template<typename S>
struct A {
template<typename T> static auto default_lambda() {
return [](const T&) { return 42; };
}
template<class U = decltype(default_lambda<S>())>
U func(U u = default_lambda<S>()) { return u; }
};
int run2 = A<double>{}.func()(3.14);
int run3 = A<char>{}.func()('a');
This patch allows the code using the same trickery that was used to allow the code in non-member functions at namespace scope.
Please see http://llvm-reviews.chandlerc.com/D1844 for richard's approval.
llvm-svn: 192166
Summary:
Operator new, new[], delete, and delete[] are all implicitly static when
declared inside a record. CXXMethodDecl already knows this, but we need
to account for that before we pick the calling convention for the
function type.
Fixes PR17371.
Reviewers: rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1761
llvm-svn: 192150
This change doesn't go all the way to making fields redeclarable; instead, it
makes them 'mergeable', which means we can find the canonical declaration, but
not much else (and for a declaration that's not from a module, the canonical
declaration is always that declaration).
llvm-svn: 192092
When nested C++11 lambdas are used in NSDMI's - this patch prevents infinite recursion by computing the linkage of any nested lambda by determining the linkage of the outermost enclosing lambda (which might inherit its linkage from its parent).
See http://llvm-reviews.chandlerc.com/D1783 for Doug's approval.
[On a related note, I need this patch so as to pass tests of transformations of nested lambdas returned from member functions]
llvm-svn: 191727
Currently, IR generation can't handle file-scope compound literals with
non-constant initializers in C++.
Fixes PR17415 (the first crash in the bug).
(We should probably change (T){1,2,3} to use the same codepath as T{1,2,3} in
C++ eventually, given that the semantics of the latter are actually defined by
the standard.)
llvm-svn: 191719
When nested lambdas are used in NSDMI's - this prevents infinite recursion.
See http://llvm-reviews.chandlerc.com/D1783 for Doug's approval regarding the code, and then request for some tests.
[On a related note, I need this patch so as to pass tests of transformations of nested lambdas returned from member functions]
llvm-svn: 191645
The general strategy is to create template versions of the conversion function and static invoker and then during template argument deduction of the conversion function, create the corresponding call-operator and static invoker specializations, and when the conversion function is marked referenced generate the body of the conversion function using the corresponding static-invoker specialization. Similarly, Codegen does something similar - when asked to emit the IR for a specialized static invoker of a generic lambda, it forwards emission to the corresponding call operator.
This patch has been reviewed in person both by Doug and Richard. Richard gave me the LGTM.
A few minor changes:
- per Richard's request i added a simple check to gracefully inform that captures (init, explicit or default) have not been added to generic lambdas just yet (instead of the assertion violation).
- I removed a few lines of code that added the call operators instantiated parameters to the currentinstantiationscope. Not only did it not handle parameter packs, but it is more relevant in the patch for nested lambdas which will follow this one, and fix that problem more comprehensively.
- Doug had commented that the original implementation strategy of using the TypeSourceInfo of the call operator to create the static-invoker was flawed and allowed const as a member qualifier to creep into the type of the static-invoker. I currently kludge around it - but after my initial discussion with Doug, with a follow up session with Richard, I have added a FIXME so that a more elegant solution that involves the use of TrivialTypeSourceInfo call followed by the correct wiring of the template parameters to the functionprototypeloc is forthcoming.
Thanks!
llvm-svn: 191634
- We scan for whitespace between comments anyways, remember any newlines seen
along the way.
- Use this newline number to decide whether two comments are adjacent.
- Since the newline check is now free remove the caching and unused code.
- Remove unnecessary boolean state from the comment list.
- No behavioral change.
llvm-svn: 191614
We previously handled one-dimensional arrays but didn't consider the
general case. The fix is simple: keep going through subsequent
dimensions until we get to the base element.
llvm-svn: 191493
The intent of getTypeOperand() was to yield an unqualified type.
However QualType::getUnqualifiedType() does not strip away qualifiers on
arrays.
N.B. This worked fine when typeid() was applied to an expression
because we would inject as implicit cast to the unqualified array type
in the AST.
llvm-svn: 191487
Specifically, the following features are not included in this commit:
- any sort of capturing within generic lambdas
- generic lambdas within template functions and nested
within other generic lambdas
- conversion operator for captureless lambdas
- ensuring all visitors are generic lambda aware
(Although I have gotten some useful feedback on my patches of the above and will be incorporating that as I submit those patches for commit)
As an example of what compiles through this commit:
template <class F1, class F2>
struct overload : F1, F2 {
using F1::operator();
using F2::operator();
overload(F1 f1, F2 f2) : F1(f1), F2(f2) { }
};
auto Recursive = [](auto Self, auto h, auto ... rest) {
return 1 + Self(Self, rest...);
};
auto Base = [](auto Self, auto h) {
return 1;
};
overload<decltype(Base), decltype(Recursive)> O(Base, Recursive);
int num_params = O(O, 5, 3, "abc", 3.14, 'a');
Please see attached tests for more examples.
This patch has been reviewed by Doug and Richard. Minor changes (non-functionality affecting) have been made since both of them formally looked at it, but the changes involve removal of supernumerary return type deduction changes (since they are now redundant, with richard having committed a recent patch to address return type deduction for C++11 lambdas using C++14 semantics).
Some implementation notes:
- Add a new Declarator context => LambdaExprParameterContext to
clang::Declarator to allow the use of 'auto' in declaring generic
lambda parameters
- Add various helpers to CXXRecordDecl to facilitate identifying
and querying a closure class
- LambdaScopeInfo (which maintains the current lambda's Sema state)
was augmented to house the current depth of the template being
parsed (id est the Parser calls Sema::RecordParsingTemplateParameterDepth)
so that SemaType.cpp::ConvertDeclSpecToType may use it to immediately
generate a template-parameter-type when 'auto' is parsed in a generic
lambda parameter context. (i.e we do NOT use AutoType deduced to
a template parameter type - Richard seemed ok with this approach).
We encode that this template type was generated from an auto by simply
adding $auto to the name which can be used for better diagnostics if needed.
- SemaLambda.h was added to hold some common lambda utility
functions (this file is likely to grow ...)
- Teach Sema::ActOnStartOfFunctionDef to check whether it
is being called to instantiate a generic lambda's call
operator, and if so, push an appropriately prepared
LambdaScopeInfo object on the stack.
- various tests were added - but much more will be needed.
There is obviously more work to be done, and both Richard (weakly) and Doug (strongly)
have requested that LambdaExpr be removed form the CXXRecordDecl LambdaDefinitionaData
in a future patch which is forthcoming.
A greatful thanks to all reviewers including Eli Friedman, James Dennett,
and especially the two gracious wizards (Richard Smith and Doug Gregor)
who spent hours providing feedback (in person in Chicago and on the mailing lists).
And yet I am certain that I have allowed unidentified bugs to creep in; bugs, that I will do my best to slay, once identified!
Thanks!
llvm-svn: 191453
1. Fixed constructor of shared clause.
2. Some macros for clauses processing are replaced by private template methods.
3. Additional checks in sema analysis of OpenMP clauses.
llvm-svn: 191265
LLVM supports applying conversion instructions to vectors of the same number of
elements (fptrunc, fptosi, etc.) but there had been no way for a Clang user to
cause such instructions to be generated when using builtin vector types.
C-style casting on vectors is already defined in terms of bitcasts, and so
cannot be used for these conversions as well (without leading to a very
confusing set of semantics). As a result, this adds a __builtin_convertvector
intrinsic (patterned after the OpenCL __builtin_astype intrinsic). This is
intended to aid the creation of vector intrinsic headers that create generic IR
instead of target-dependent intrinsics (in other words, this is a generic
_mm_cvtepi32_ps). As noted in the documentation, the action of
__builtin_convertvector is defined in terms of the action of a C-style cast on
each vector element.
llvm-svn: 190915
Summary:
When selecting a mangling for an anonymous tag type:
- We should first try it's typedef'd name.
- If that doesn't work, we should mangle in the name of the declarator
that specified it as a declaration specifier.
- If that doesn't work, fall back to a static mangling of
<unnamed-type>.
This should make our anonymous type mangling compatible.
This partially fixes PR16994; we would need to have an implementation of
scope numbering to get it right (a separate issue).
Reviewers: rnk, rsmith, rjmccall, cdavis5x
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1540
llvm-svn: 190892
Like any other type, an init list for a vector can have the same type as
the vector itself; handle that case.
<rdar://problem/14990460>
llvm-svn: 190844
Summary:
This fixes several issues with the original implementation:
- Win32 entry points cannot be in namespaces
- A Win32 entry point cannot be a function template, diagnose if we it.
- Win32 entry points cannot be overloaded.
- Win32 entry points implicitly return, similar to main.
Reviewers: rnk, rsmith, whunt, timurrrr
Reviewed By: rnk
CC: cfe-commits, nrieck
Differential Revision: http://llvm-reviews.chandlerc.com/D1683
llvm-svn: 190818
address spaces which is both (1) a "semantic" concept and
(2) possibly a hardware level restriction. It is desirable to
be able to discard/merge the LLVM-level address spaces on arguments for which
there is no difference to the current backend while keeping
track of the semantic address spaces in a funciton prototype. To do this
enable addition of the address space into the name-mangling process. Add
some tests to document this behaviour against inadvertent changes.
Patch by Michele Scandale!
llvm-svn: 190684
Summary:
Functions named "main", "wmain", "WinMain", "wWinMain", and "DllMain"
are never mangled regardless of linkage, even when compiling for kernel
mode.
Depends on D1655
Reviewers: timurrrr, pcc, rnk, whunt
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1670
llvm-svn: 190675
Summary:
This is a first step to getting extern "C" working properly inside
clang. There are a number of quirks but mangling declarations inside
such a function are a good first step.
Reviewers: timurrrr, pcc, cdavis5x
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1655
llvm-svn: 190671
Summary:
More accurately characterize the nature of array parameters. Doing this
removes false back-reference opportunities. Remove some hacks now that
we characterize these better.
Reviewers: rnk, timurrrr, whunt, cdavis5x
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1626
llvm-svn: 190488
trunk clang is a bit more aggressive about emitting unused-declaration
warnings, so adjust some AST code to match. Specifically, use
LLVM_ATTRIBUTE_UNUSED for declarations which are never supposed to be
referenced, and turn references to declarations which are supposed to be
referenced into odr-uses.
llvm-svn: 190443
Static locals requiring initialization are not thread safe on Windows.
Unfortunately, it's possible to create static locals that are actually
externally visible with inline functions and templates. As a result, we
have to implement an initialization guard scheme that is compatible with
TUs built by MSVC, which makes thread safety prohibitively difficult.
MSVC's scheme is that every function that requires a guard gets an i32
bitfield. Each static local is assigned a bit that indicates if it has
been initialized, up to 32 bits, at which point a new bitfield is
created. MSVC rejects inline functions with more than 32 static locals,
and the externally visible mangling (?_B) only allows for one guard
variable per function.
On Eli's recommendation, I used MangleNumberingContext to track which
bit each static corresponds to.
Implements PR16888.
Reviewers: rjmccall, eli.friedman
Differential Revision: http://llvm-reviews.chandlerc.com/D1416
llvm-svn: 190427
name lookup from lazily deserializing the other declarations with the same
name, by tracking a bit to indicate whether a name in a DeclContext might have
additional external results. This also allows lazier reconciling of the lookup
table if a module import adds decls to a pre-existing DC.
However, this exposes a pre-existing bug, which causes a regression in
test/Modules/decldef.mm: if we have a reference to a declaration, and a
later-imported module adds a redeclaration, nothing causes us to load that
redeclaration when we use or emit the reference (which can manifest as a
reference to an undefined inline function, a use of an incomplete type, and so
on). decldef.mm has been extended with an additional testcase which fails with
or without this change.
llvm-svn: 190293
Summary:
__uuidof on templated types should exmaine if any of its template
parameters have a uuid declspec. If exactly one does, then take it.
Otherwise, issue an appropriate error.
Reviewers: rsmith, thakis, rnk
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1419
llvm-svn: 190240
Summary: Closure classes for C++ lambdas are always compiler-generated. This one-line change calls setImplicit(true) on them at creation time, such that a default RecursiveASTVisitor (or any for which shouldVisitImplicitCode returns false) will skip them.
Reviewers: rsmith, dblaikie
Reviewed By: dblaikie
CC: klimek, revane, cfe-commits, jordan_rose
Differential Revision: http://llvm-reviews.chandlerc.com/D1593
llvm-svn: 190073
getRealTypeByWidth and getIntTypeByWidth
for ASTContext names are almost same(invokes new methods from TargetInfo):
getIntTypeForBitwidth and getRealTypeForBitwidth.
As first commit for PR16752 fix: 'mode' attribute for unusual targets doesn't work properly
Description:
Troubles could be happened due to some assumptions in handleModeAttr function (see SemaDeclAttr.cpp).
For example, it assumes that 32 bit integer is 'int', while it could be 16 bit only.
Instead of asking target: 'which type do you want to use for int32_t ?' it just hardcodes general opinion. That doesn't looks pretty correct.
Please consider the next solution:
1. In Basic/TargetInfo add getIntTypeByWidth and getRealTypeByWidth virtual methods. By default current behaviour could be implemented here.
2. Fix handleModeAttr according to new methods in TargetInfo.
This approach is implemented in the patch attached to this post.
Fixes:
1st Commit (Current): Add new methods for TargetInfo:
getRealTypeByWidth and getIntTypeByWidth
for ASTContext names are almost same(invokes new methods from TargetInfo):
getIntTypeForBitwidth and getRealTypeForBitwidth
2nd Commit (Next): Fix SemaDeclAttr, handleModeAttr function.
llvm-svn: 190044
When an AST file is built based on another AST file, it can use a decl from
the fist file, and therefore mark the "isUsed" bit. We need to note this in
the AST file so that the bit is set correctly when the second AST file is
loaded.
This patch introduces the distinction between setIsUsed() and markUsed() so
that we don't call into the ASTMutationListener callback when it wouldn't
be appropriate.
Fixes PR16635.
llvm-svn: 190016
Summary: I added the display of the VarDecl contained in the statement.
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1596
llvm-svn: 189941
Summary:
I have no idea why these were there in the first place, but now they are
certainly not necessary.
Reviewers: rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1581
llvm-svn: 189813
been an oversight, as it definitely works. Every test which changed had
the const written on the LHS of the auto already.
Notably, this also makes things like cpp11-migrate's formation of 'const
auto &' variables much more familiar.
Yes, many people feel that 'const' and other qualifiers belong on the
RHS of the type. I'm not going to argue about that because Clang already
*overwhelming* places the qualifiers on the LHS when it can and on the
RHS when it must. We shouldn't diverge for auto. We should add a tool to
clang-tidy that fixes this in either direction, and then wire up
clang-tidy to tools like cpp11-migrate to fix their placement after
transforms.
llvm-svn: 189769
these in eagerly if we're not actually processing a translation unit. The added
laziness here also avoids us loading in parts of a CXXRecordDecl earlier than an
upcoming class template specialization merging patch would like.
Ideally, we should mark the vtable as used when we see a definition for the key
function, rather than having a separate pass over dynamic classes at the end of
the TU. The existing approach is pretty bad for PCH/modules, since it forcibly
loads the declarations of all key functions in all imported modules, whether or
not those key functions are defined.
llvm-svn: 189627
I changed the diagnostic printing code because it's probably better
to cut off a digit from DBL_MAX than to print something like
1.300000001 when the user wrote 1.3.
llvm-svn: 189625
Summary:
Instead of calling getAsTemplate(), call
getAsTemplateOrTemplatePattern() because it handles the
TemplateExpansion case too.
This fixes PR16997.
Reviewers: doug.gregor, rsmith
Reviewed By: rsmith
CC: cfe-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D1512
llvm-svn: 189422
Summary:
Makes functions with implicit calling convention compatible with
function types with a matching explicit calling convention. This fixes
things like calls to qsort(), which has an explicit __cdecl attribute on
the comparator in Windows headers.
Clang will now infer the calling convention from the declarator. There
are two cases when the CC must be adjusted during redeclaration:
1. When defining a non-inline static method.
2. When redeclaring a function with an implicit or mismatched
convention.
Fixes PR13457, and allows clang to compile CommandLine.cpp for the
Microsoft C++ ABI.
Excellent test cases provided by Alexander Zinenko!
Reviewers: rsmith
Differential Revision: http://llvm-reviews.chandlerc.com/D1231
llvm-svn: 189412
- __func__ or __FUNCTION__ returns captured statement's parent
function name, not the one compiler generated.
Differential Revision: http://llvm-reviews.chandlerc.com/D1491
Reviewed by bkramer
llvm-svn: 189219
would cause us to concatenate these paragraphs into a single one.
The no-op whitespace churn in test/Index test happened because these tests
don't use the correct approach for testing and are more strict than required
for they are testing.
llvm-svn: 189126
This was only used to ensure that the traversal order was the same as the
insertion order, but that guarantee was already being provided by the use
of a FoldingSetVector.
llvm-svn: 189075
Richard Smith