of decl bit offsets.
This allows us to easily get at the location of a decl without deserializing it.
It increases size of Cocoa PCH by only 0.2%.
llvm-svn: 143123
AST file more lazy, so that we don't eagerly load that information for
all known identifiers each time a new AST file is loaded. The eager
reloading made some sense in the context of precompiled headers, since
very few identifiers were defined before PCH load time. With modules,
however, a huge amount of code can get parsed before we see an
@import, so laziness becomes important here.
The approach taken to make this information lazy is fairly simple:
when we load a new AST file, we mark all of the existing identifiers
as being out-of-date. Whenever we want to access information that may
come from an AST (e.g., whether the identifier has a macro definition,
or what top-level declarations have that name), we check the
out-of-date bit and, if it's set, ask the AST reader to update the
IdentifierInfo from the AST files. The update is a merge, and we now
take care to merge declarations before/after imports with declarations
from multiple imports.
The results of this optimization are fairly dramatic. On a small
application that brings in 14 non-trivial modules, this takes modules
from being > 3x slower than a "perfect" PCH file down to 30% slower
for a full rebuild. A partial rebuild (where the PCH file or modules
can be re-used) is down to 7% slower. Making the PCH file just a
little imperfect (e.g., adding two smallish modules used by a bunch of
.m files that aren't in the PCH file) tips the scales in favor of the
modules approach, with 24% faster partial rebuilds.
This is just a first step; the lazy scheme could possibly be improved
by adding versioning, so we don't search into modules we already
searched. Moreover, we'll need similar lazy schemes for all of the
other lookup data structures, such as DeclContexts.
llvm-svn: 143100
essence, the redeclaration chain for a class could end up in an
inconsistent state while deserializing multiple declarations in that
chain, where the circular linked list was not, in fact,
circular. Since only two redeclarations of the same entity will get
loaded when we're in this state, restore circularity when both have
been loaded. Fixes <rdar://problem/10324940> / PR11195.
llvm-svn: 143037
expressions: expressions which refer to a logical rather
than a physical l-value, where the logical object is
actually accessed via custom getter/setter code.
A subsequent patch will generalize the AST for these
so that arbitrary "implementing" sub-expressions can
be provided.
Right now the only client is ObjC properties, but
this should be generalizable to similar language
features, e.g. Managed C++'s __property methods.
llvm-svn: 142914
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
preprocessed entities that are #included in the range that we are interested.
This is useful when we are interested in preprocessed entities of a specific file, e.g
when we are annotating tokens. There is also an optimization where we cache the last
result of PreprocessingRecord::getPreprocessedEntitiesInRange and we re-use it if
there is a call with the same range as before.
rdar://10313365
llvm-svn: 142887
in such a case just write out a reference of a previously serialized Stmt, instead
of serializing it all over again.
This saves memory + space + [de]serializing time, and avoids blowing up memory
with pathological cases. rdar://10293911
llvm-svn: 142696
-Add the location of the class name to all objc container decls, not just ObjCInterfaceDecl.
-Make objc decls consistent with the rest of the NamedDecls and have getLocation() point to the
class name, not the location of '@'.
llvm-svn: 141061
Instead of always storing all source locations for the selector identifiers
we check whether all the identifiers are in a "standard" position; "standard" position is
-Immediately before the arguments: -(id)first:(int)x second:(int)y;
-With a space between the arguments: -(id)first: (int)x second: (int)y;
-For nullary selectors, immediately before ';': -(void)release;
In such cases we infer the locations instead of storing them.
llvm-svn: 140989
Instead of always storing all source locations for the selector identifiers
we check whether all the identifiers are in a "standard" position; "standard" position is
-Immediately before the arguments: [foo first:1 second:2]
-With a space between the arguments: [foo first: 1 second: 2]
-For nullary selectors, immediately before ']': [foo release]
In such cases we infer the locations instead of storing them.
llvm-svn: 140987
PreprocessingRecord's getPreprocessedEntitiesInRange.
Also remove all the stuff that were added in ASTUnit that are unnecessary now
that we do a binary search for preprocessed entities and deserialize only
what is necessary.
llvm-svn: 140063
check whether the requested location points inside the precompiled preamble,
in which case the returned source location will be a "loaded" one.
llvm-svn: 140060
which will do a binary search and return a pair of iterators
for preprocessed entities in the given source range.
Source ranges of preprocessed entities are stored twice currently in
the PCH/Module file but this will be fixed in a subsequent commit.
llvm-svn: 140058
don't call ReadSLocEntryRecord() directly because the entry may have
already been loaded in which case calling ReadSLocEntryRecord()
directly would trigger an assertion in SourceManager.
llvm-svn: 140052
arbitrary amount of code. This forces us to stage the AST writer more
strictly, ensuring that we don't assign a declaration ID to a
declaration until after we're certain that no more modules will get
loaded.
llvm-svn: 139974
-Use an array of offsets for all preprocessed entities
-Get rid of the separate array of offsets for just macro definitions;
for references to macro definitions use an index inside the preprocessed
entities array.
-Deserialize each preprocessed entity lazily, at first request; not in bulk.
Paves the way for binary searching of preprocessed entities that will offer
efficiency and will simplify things on the libclang side a lot.
llvm-svn: 139809
to the consumer without being fully deserialized).
The regression was on compiling boost.python and it was too difficult to get a reduced
test case unfortunately.
Also modify the logic of how objc methods are getting passed to the consumer;
codegen depended on receiving objc methods before the implementation decl.
Since the interesting objc methods are ones with a body and such methods only
exist inside an ObjCImplDecl, deserialize and pass to consumer all the methods
of ObCImplDecl when we see one.
Fixes http://llvm.org/PR10922 & rdar://10117105.
llvm-svn: 139644
language options. Use that .def file to declare the LangOptions class
and initialize all of its members, eliminating a source of annoying
initialization bugs.
AST serialization changes are next up.
llvm-svn: 139605
modifying directly for the preamble.
This avoids an awful, hard to find, bug where "PreprocessorOpts.DisablePCHValidation = true"
would be persistent for subsequent reparses of the translation unit which would result
in defines, present in command-line but not in the PCH, being ignored.
Fixes rdar://9615399.
llvm-svn: 139512
than having CodeGen check whether a declaration comes from an AST file
(which it shouldn't know or care about), make sure that the AST writer and
reader pass along "interesting" declarations that CodeGen needs to
know about.
llvm-svn: 139441
declaration was deserialized from an AST file. Use this instead of
Decl::getPCHLevel() wherever possible. This is a simple step toward
killing off Decl::getPCHLevel().
llvm-svn: 139427
ASTContext reference. Remove all of the extra checking and logic that
was used to cope with a NULL ASTContext. No effective functionality
change.
llvm-svn: 139413
identifier, also make them visible in the translation unit," which
isn't needed now that John's eliminated the AST dependency in blocks
CodeGen.
llvm-svn: 139408
'id' that can be used (only!) via a contextual keyword as the result
type of an Objective-C message send. 'instancetype' then gives the
method a related result type, which we have already been inferring for
a variety of methods (new, alloc, init, self, retain). Addresses
<rdar://problem/9267640>.
llvm-svn: 139275
Preprocessor, eliminating the constructor that was used by ASTUnit
(which didn't provide an ASTContext or Prepreprocessor). Ensuring that
both objects are non-NULL will simplify module loading (but none of
that is done yet).
llvm-svn: 138986
builtin types (When requested). This is another step toward making
ASTUnit build the ASTContext as needed when loading an AST file,
rather than doing so after the fact. No actual functionality change (yet).
llvm-svn: 138985
include guards don't show up as macro definitions in every translation
unit that imports a module. Macro definitions can, however, be
exported with the intentionally-ugly #__export_macro__
directive. Implement this feature by not even bothering to serialize
non-exported macros to a module, because clients of that module need
not (should not) know that these macros even exist.
llvm-svn: 138943
The initial incentive was to fix a crash when PCH chaining categories
to an interface, but the fix was done in the "modules way" that I hear
is popular with the kids these days.
Each module stores the local chain of categories and we combine them
when the interface is loaded. We also warn if non-dependent modules
introduce duplicate named categories.
llvm-svn: 138926
sure that all of the CXXConversionDecls go into the same
bucket. Otherwise, name lookup might not find them all. Fixes
<rdar://problem/10041960>.
llvm-svn: 138824
Empty lookups can occur in the DeclContext map when we are chaining PCHs, where
the empty lookup indicates that we already looked in ExternalASTSource.
llvm-svn: 138816
, such as list of forward @class decls, in a DeclGroup
node. Deal with its consequence throught clang. This
is in preparation for more Sema work ahead. // rdar://8843851.
Feel free to reverse if it breaks something important
and I am unavailable.
llvm-svn: 138709
from the given source. -emit-module behaves similarly to -emit-pch,
except that Sema is somewhat more strict about the contents of
-emit-module. In the future, there are likely to be more interesting
differences.
llvm-svn: 138595
to "when loading a particular module" validation, since it was only
validating local information anyway. This shouldn't change anything.
llvm-svn: 138583
redeclarations of a particular entity would occur in source
order. Friend declarations that occur within class templates (or
member classes thereof) do not follow this, nor would modules. Big
thanks to Erik Verbruggen for reducing this problem from the Very
Large Qt preamble testcase he found.
llvm-svn: 138557
given selector, rather than walking the chain backwards. Teach its
visitor how to merge multiple result sets into a single result set,
combining the results of selector lookup in several different modules
into a single result set.
llvm-svn: 138556
table when serializing an AST file. This was a holdover from the days
before chained PCH, and is a complete waste of time and storage
now. It's a good thing it's useless, because I have no idea how I
would have implemented MaterializeVisibleDecls efficiently in the
presence of modules.
llvm-svn: 138496
which supports both pre-order and post-order traversal via a visitor
mechanism. Use this depth-first search with a post-order traversal to
give predictable ordering semantics when walking all of the lexical
declarations in the translation unit.
Eventually, module imports will occur in the source code rather than
at the beginning, and we'll have to revisit this walk.
llvm-svn: 138490
module DAG-based lookup scheme. This required some reshuffling, so
that each module stores its own mapping from DeclContexts to their
lexical and visible sets for those DeclContexts (rather than one big
"chain").
Overall, this allows simple qualified name lookup into the translation
unit to gather results from multiple modules, with the lookup results
in module B shadowing the lookup results in module A when B imports A.
Walking all of the lexical declarations in a module DAG is still a
mess; we'll end up walking the loaded module list backwards, which
works fine for chained PCH but doesn't make sense in a DAG. I'll
tackle this issue as a separate commit.
llvm-svn: 138463
Currently getMacroArgExpandedLocation is very inefficient and for the case
of a location pointing at the main file it will end up checking almost all of
the SLocEntries. Make it faster:
-Use a map of macro argument chunks to their expanded source location. The map
is for a single source file, it's stored in the file's ContentCache and lazily
computed, like the source lines cache.
-In SLocEntry's FileInfo add an 'unsigned NumCreatedFIDs' field that keeps track
of the number of FileIDs (files and macros) that were created during preprocessing
of that particular file SLocEntry. This is useful when computing the macro argument
map in skipping included files while scanning for macro arg FileIDs that lexed from
a specific source file. Due to padding, the new field does not increase the size
of SLocEntry.
llvm-svn: 138225
different modules) more robust. It already handled (simple) merges of
the set of declarations attached to that identifier, so add a test
case that shows us getting two different declarations for the same
identifier (one struct, one function) from different modules, and are
able to use both of them.
llvm-svn: 138189
modules (those that no other module depends on) and performs a search
over all of the modules, visiting a new module only when all of the
modules that depend on it have already been visited. The visitor can
abort the search for all modules that a module depends on, which
allows us to minimize the number of lookups necessary when performing
a search.
Switch identifier lookup from a linear walk over the set of modules to
this module visitation operation. The behavior is the same for simple
PCH and chained PCH, but provides the proper search order for
modules. Verified with printf debugging, since we don't have enough in
place to actually test this.
llvm-svn: 138187
has already been loaded before allocating a new Module structure. If
the module has already been loaded (uniquing based on file name), then
just return the existing module rather than trying to load it again.
This allows us to load a DAG of modules. Introduce a simple test case
that forms a diamond-shaped module graph, and illustrates that a
source file importing the bottom of the diamond can see declarations
in all four of the modules that make up the diamond.
Note that this version moves the file-opening logic into the module
manager, rather than splitting it between the module manager and the
AST reader. More importantly, it properly handles the
weird-but-possibly-useful case of loading an AST file from "-".
llvm-svn: 138030
Teach ModuleManager::addModule() to check whether a particular module
has already been loaded before allocating a new Module structure. If
the module has already been loaded (uniquing based on file name), then
just return the existing module rather than trying to load it again.
This allows us to load a DAG of modules. Introduce a simple test case
that forms a diamond-shaped module graph, and illustrates that a
source file importing the bottom of the diamond can see declarations
in all four of the modules that make up the diamond.
llvm-svn: 137971
has already been loaded before allocating a new Module structure. If
the module has already been loaded (uniquing based on file name), then
just return the existing module rather than trying to load it again.
This allows us to load a DAG of modules. Introduce a simple test case
that forms a diamond-shaped module graph, and illustrates that a
source file importing the bottom of the diamond can see declarations
in all four of the modules that make up the diamond.
llvm-svn: 137925
-import-module) vs. loaded because some other module depends on
them. As part of doing this, pass down the module that caused a module
to be loaded directly, rather than assuming that we're loading a
chain. Finally, write out all of the directly-loaded modules when
serializing an AST file (using the new IMPORTS record), so that an AST
file can depend on more than one other AST file, all of which will be
loaded when that AST file is loaded. This allows us to form and load a
tree of modules, but we can't yet load a DAG of modules.
llvm-svn: 137923
all AST files have a normal METADATA record that has the same form
regardless of whether we refer to a chained PCH or any other kind of
AST file.
Introduce the IMPORTS record, which describes all of the AST files
that are imported by this AST file, and how (as a module, a PCH file,
etc.). Currently, we emit at most one entry to this record, to support
chained PCH.
llvm-svn: 137869
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
type over into the AST context, then make that declaration a
predefined declaration in the AST format. This ensures that different
AST files will at least agree on the (global) declaration ID for 'id',
and eliminates one of the "special" types in the AST file format.
llvm-svn: 137429
eliminating a pile of redundant code (and probably some bugs in the
process). The variation between chained and non-chained PCH is fairly
small now anyway.
llvm-svn: 137410
declaration that never actually gets serialized. Instead, serialize
the various kinds of update records (lexical decls, visible decls, the
addition of an anonymous namespace) for the translation unit, even if
we're not chaining. This way, we won't have to deal with multiple
loaded translation unit declarations.
llvm-svn: 137395
either "special" type has already been initialized. Previously, we did
this check based on just the first special type (__builtin_va_list),
but now we have some NULL special type entries to content with.
llvm-svn: 137373
enumerations from the ASTContext into CodeGen, so that we don't need
to serialize it to AST files. This appears to be the last of the
low-hanging fruit for SpecialTypes.
llvm-svn: 137124
layout of a constant NSString from the ASTContext over to CodeGen,
since this is solely CodeGen's responsibility. Eliminates one of the
unnecessary "special" types that we serialize.
llvm-svn: 137121
the last of the ID/offset/index mappings that I know
of. Unfortunately, the "gap" method of testing doesn't work here due
to the way the preprocessing record performs iteration. We'll do more
testing once multi-AST loading is possible.
llvm-svn: 136902
IDs will never cross module boundaries, since they're tied to the
CXXDefinitionData, so just use a local mapping throughout. Eliminate
the global -> local tables and supporting data.
llvm-svn: 136847
AST file, along with an enumeration naming those predefined
declarations. No functionality change, but this will make it easier to
introduce new predefined declarations, when/if we need them.
llvm-svn: 136781
reader, to allow AST files to be loaded with their declarations
remapped to different ID numbers. Fix a number of places where we were
either failing to map local declaration IDs into global declaration
IDs or where interpreting the local declaration IDs within the wrong
module.
I've tested this via the usual "random gaps" method. It works well
except for the preamble tests, because our handling of the precompiled
preamble requires declaration and preprocessed entity to be stable
when parsing code and then loading that back into memory. This
property will hold in general, but my randomized testing naturally
breaks this property to get more coverage. In the future, I expect
that the precompiled preamble logic won't need this property.
I am very unhappy with the current handling of the translation unit,
which is a rather egregious hack. We're going to have to do something
very different here for loading multiple AST files, because we don't
want to have to cope with merging two translation units. Likely, we'll
just handle translation units entirely via "update" records, and
predefine a single, fixed declaration ID for the translation
unit. That will come later.
llvm-svn: 136779
by eliminating the type ID from constructor, destructor, and
conversion function names. There are several reasons for this change:
- A given type (say, int*) isn't guaranteed to have a single, unique
type ID within a chain of PCH files. Hence, we could end up hashing
based on the wrong type ID, causing name lookup to fail.
- The mapping from types back to type IDs required one DenseMap
entry for every type that was ever deserialized, which was an
unacceptable cost to support just the name lookup of constructors,
destructors, and conversion functions. Plus, this mapping could
never actually work with chained or multiple PCH, based on the first
bullet.
Once we have eliminated the type from the hash function, these
problems go away, as does my horrible "reverse type remap" hack, which
was doomed from the start (see bullet #1 above) and far too
complicated.
However, note that removing the type from the hash function means that
all constructors, destructors, and conversion functions have the same
hash key, so I've updated the caller to double-check that the
declarations found have the appropriate name.
llvm-svn: 136708
reader. This scheme permits an AST file to be loaded with its type IDs
shifted anywhere in the type ID space.
At present, the type indices are still allocated in the same boring
way they always have been, just by adding up the number of types in
each PCH file within the chain. However, I've done testing with this
patch by randomly sliding the base indices at load time, to ensure
that remapping is occurring as expected. I may eventually formalize
this in some testing flag, but loading multiple (non-chained) AST
files at once will eventually exercise the same code.
There is one known problem with this patch, which involves name lookup
of operator names (e.g., "x.operator int*()") in cases where multiple
PCH files in the chain. The hash function itself depends on having a
stable type ID, which doesn't happen with chained PCH and *certainly*
doesn't happen when sliding type IDs around. We'll need another
approach. I'll tackle that next.
llvm-svn: 136693
reader statistics), to show the local-to-global mappings. The only
such mapping we have (at least, for now) is for source location
offsets.
llvm-svn: 136687
were (Module*, Offset) with equivalent maps whose value type is just a
Module*. The offsets have moved into corresponding "Base" fields
within the Module itself, where they will also be helpful for
local->global translation (eventually).
llvm-svn: 136441
point, ASTReader::InitializeSema() has very little interesting work,
*except* issues stemming from preloaded declarations. That's something
we'll still need to cope with.
llvm-svn: 136378
completely broken deserialization mapping code we had for VTableUses,
which would have broken horribly as soon as our local-to-global ID
mapping became interesting.
llvm-svn: 136371
we could turn this into an on-disk hash table so we don't load the
whole thing the first time we need it. However, it tends to be very,
very small (i.e., empty) for most precompiled headers, so it isn't all
that interesting.
llvm-svn: 136352
- Added LazyVector::erase() to support this use case.
- Factored out the LazyDecl-of-Decls to RecordData translation in
the ASTWriter. There is still a pile of code duplication here to
eliminate.
llvm-svn: 136270
contents are lazily loaded on demand from an external source (e.g., an
ExternalASTSource or ExternalSemaSource). The "loaded" entities are
kept separate from the "local" entities, so that the two can grow
independently.
Switch Sema::TentativeDefinitions from a normal vector that is eagerly
populated by the ASTReader into one of these LazyVectors, making the
ASTReader a bit more like me (i.e., lazy).
llvm-svn: 136262
etc. With this I think essentially all of the SourceManager APIs are
converted. Comments and random other bits of cleanup should be all thats
left.
llvm-svn: 136057
and various other 'expansion' based terms. I've tried to reformat where
appropriate and catch as many references in comments but I'm going to do
several more passes. Also I've tried to expand parameter names to be
more clear where appropriate.
llvm-svn: 136056
so that we have one, simple way to map from global bit offsets to
local bit offsets. Eliminates a number of loops over the chain, and
generalizes for more interesting bit remappings.
Also, as an amusing oddity, we were computing global bit offsets
*backwards* for preprocessed entities (e.g., the directly included PCH
file in the chain would start at offset zero, rather than the original
PCH that occurs first in translation unit). Even more amusingly, it
made precompiled preambles work, because we were forgetting to adjust
the local bit offset to a global bit offset when storing preprocessed
entity offsets in the ASTUnit. Two wrongs made a right, and now
they're both right.
llvm-svn: 135750
type IDs into a single place, and make sure that all of the callers
use the appropriate functions to do the mapping. Since the mapping is
still the identity function, this is essentially a no-op.
llvm-svn: 135733
such that every declaration ID loaded from an AST file will go through
a central local -> global mapping function. At present, this change
does nothing, since the local -> global mapping function is the
identity function.
This is the mechanical part of the refactoring; a follow-up patch will
address a few remaining areas where it's not obvious whether we're
dealing with local or global IDs.
llvm-svn: 135711
entities generated directly by the preprocessor from those loaded from
the external source (e.g., the ASTReader). By separating these two
sets of entities into different vectors, we allow both to grow
independently, and eliminate the need for preallocating all of the
loaded preprocessing entities. This is similar to the way the recent
SourceManager refactoring treats FileIDs and the source location
address space.
As part of this, switch over to building a continuous range map to
track preprocessing entities.
llvm-svn: 135646
the AST reader down to the AST file + local ID, rather than walking
the PCH chain. More cleanup/generalization, although there is more
work to do for preprocessed entities. In particular, the
"preallocation" scheme for preprocessed entities is not going to work
well with late loading of PCH files, and it's likely we'll have to do
something akin to the SourceManager's negative/positive loading.
llvm-svn: 135556
reader down to the AST file + local ID, rather than walking the PCH
chain. No functionality change; this is generalization and cleanup.
llvm-svn: 135554
AST reader down to the AST file + local ID, rather than walking the
PCH chain. No functionality change; this is generalization and cleanup.
llvm-svn: 135551
AST reader down to the AST file + local ID within that file, rather
than lamely walking the PCH chain. There's no actual functionality
change now, but this is cleaner and more general.
llvm-svn: 135548
source locations from source locations loaded from an AST/PCH file.
Previously, loading an AST/PCH file involved carefully pre-allocating
space at the beginning of the source manager for the source locations
and FileIDs that correspond to the prefix, and then appending the
source locations/FileIDs used for parsing the remaining translation
unit. This design forced us into loading PCH files early, as a prefix,
whic has become a rather significant limitation.
This patch splits the SourceManager space into two parts: for source
location "addresses", the lower values (growing upward) are used to
describe parsed code, while upper values (growing downward) are used
for source locations loaded from AST/PCH files. Similarly, positive
FileIDs are used to describe parsed code while negative FileIDs are
used to file/macro locations loaded from AST/PCH files. As a result,
we can load PCH/AST files even during parsing, making various
improvemnts in the future possible, e.g., teaching #include <foo.h> to
look for and load <foo.h.gch> if it happens to be already available.
This patch was originally written by Sebastian Redl, then brought
forward to the modern age by Jonathan Turner, and finally
polished/finished by me to be committed.
llvm-svn: 135484
to allow clients to specify that they've already (correctly) loaded
declarations, and that no further action is needed.
Also, make sure that we clear the "has external lexical declarations"
bit before calling FindExternalLexicalDecls(), to avoid infinite
recursion.
llvm-svn: 135306
Also add the missing serialization support for SEHTryStmt,
SEHFinallyStmt, and SEHExceptStmt, and fix and finish the
serialization support for AsTypeExpr. In addition, change
the code so that it will no longer link if a Stmt subclass
is missing serialization support.
llvm-svn: 135258
variants to 'expand'. This changed a couple of public APIs, including
one public type "MacroInstantiation" which is now "MacroExpansion". The
rest of the codebase was updated to reflect this, especially the
libclang code. Two of the C++ (and thus easily changed) libclang APIs
were updated as well because they pertained directly to the old
MacroInstantiation class.
No functionality changed.
llvm-svn: 135139
throw-expressions, such that we don't consider the NRVO when the
non-volatile automatic object comes from outside the innermost try
scope (C++0x [class.copymove]p13). In C++98/03, our ASTs were
incorrect but it didn't matter because IR generation doesn't actually
apply the NRVO here. In C++0x, however, we were moving from an object
when in fact we should have copied from it. Fixes PR10142 /
<rdar://problem/9714312>.
llvm-svn: 134548
type/expression/template argument/etc. is instantiation-dependent if
it somehow involves a template parameter, even if it doesn't meet the
requirements for the more common kinds of dependence (dependent type,
type-dependent expression, value-dependent expression).
When we see an instantiation-dependent type, we know we always need to
perform substitution into that instantiation-dependent type. This
keeps us from short-circuiting evaluation in places where we
shouldn't, and lets us properly implement C++0x [temp.type]p2.
In theory, this would also allow us to properly mangle
instantiation-dependent-but-not-dependent decltype types per the
Itanium C++ ABI, but we aren't quite there because we still mangle
based on the canonical type in cases like, e.g.,
template<unsigned> struct A { };
template<typename T>
void f(A<sizeof(sizeof(decltype(T() + T())))>) { }
template void f<int>(A<sizeof(sizeof(int))>);
and therefore get the wrong answer.
llvm-svn: 134225
for a template template parameter.
Uses to follow.
I've also made the uniquing of SubstTemplateTemplateParmPacks
use a ContextualFoldingSet as a minor space efficiency.
llvm-svn: 134137
vector<int>
to
std::vector<int>
Patch by Kaelyn Uhrain, with minor tweaks + PCH support from me. Fixes
PR5776/<rdar://problem/8652971>.
Thanks Kaelyn!
llvm-svn: 134007
MaterializeTemporaryExpr captures a reference binding to a temporary
value, making explicit that the temporary value (a prvalue) needs to
be materialized into memory so that its address can be used. The
intended AST invariant here is that a reference will always bind to a
glvalue, and MaterializeTemporaryExpr will be used to convert prvalues
into glvalues for that binding to happen. For example, given
const int& r = 1.0;
The initializer of "r" will be a MaterializeTemporaryExpr whose
subexpression is an implicit conversion from the double literal "1.0"
to an integer value.
IR generation benefits most from this new node, since it was
previously guessing (badly) when to materialize temporaries for the
purposes of reference binding. There are likely more refactoring and
cleanups we could perform there, but the introduction of
MaterializeTemporaryExpr fixes PR9565, a case where IR generation
would effectively bind a const reference directly to a bitfield in a
struct. Addresses <rdar://problem/9552231>.
llvm-svn: 133521
they should still be officially __strong for the purposes of errors,
block capture, etc. Make a new bit on variables, isARCPseudoStrong(),
and set this for 'self' and these enumeration-loop variables. Change
the code that was looking for the old patterns to look for this bit,
and change IR generation to find this bit and treat the resulting
variable as __unsafe_unretained for the purposes of init/destroy in
the two places it can come up.
llvm-svn: 133243
Language-design credit goes to a lot of people, but I particularly want
to single out Blaine Garst and Patrick Beard for their contributions.
Compiler implementation credit goes to Argyrios, Doug, Fariborz, and myself,
in no particular order.
llvm-svn: 133103
Related result types apply Cocoa conventions to the type of message
sends and property accesses to Objective-C methods that are known to
always return objects whose type is the same as the type of the
receiving class (or a subclass thereof), such as +alloc and
-init. This tightens up static type safety for Objective-C, so that we
now diagnose mistakes like this:
t.m:4:10: warning: incompatible pointer types initializing 'NSSet *'
with an
expression of type 'NSArray *' [-Wincompatible-pointer-types]
NSSet *array = [[NSArray alloc] init];
^ ~~~~~~~~~~~~~~~~~~~~~~
/System/Library/Frameworks/Foundation.framework/Headers/NSObject.h:72:1:
note:
instance method 'init' is assumed to return an instance of its
receiver
type ('NSArray *')
- (id)init;
^
It also means that we get decent type inference when writing code in
Objective-C++0x:
auto array = [[NSMutableArray alloc] initWithObjects:@"one", @"two",nil];
// ^ now infers NSMutableArray* rather than id
llvm-svn: 132868
__builtin_astype(): Used to reinterpreted as another data type of the same size using for both scalar and vector data types.
Added test case.
llvm-svn: 132612
in ASTReader::validateFileEntries().
This avoids going through all source location entries and fixes the performance regression.
Many thanks to Doug for the hint!
(rdar://9530587)
llvm-svn: 132481
a file was modified since the time the PCH was created.
The parser is not fit to deal with stale PCHs, too many invariants do not hold up. rdar://9530587.
llvm-svn: 132389
type that turns one type into another. This is used as the basis to
implement __underlying_type properly - with TypeSourceInfo and proper
behavior in the face of templates.
llvm-svn: 132017
hasTrivialDefaultConstructor() really really means it now.
Also implement a fun standards bug regarding aggregates. Doug, if you'd
like, I can un-implement that bug if you think it is truly a defect.
The bug is that non-special-member constructors are never considered
user-provided, so the following is an aggregate:
struct foo {
foo(int);
};
It's kind of bad, but the solution isn't obvious - should
struct foo {
foo (int) = delete;
};
be an aggregate or not?
Lastly, add a missing initialization to FunctionDecl.
llvm-svn: 131101
build a precompiled header. Use this information to eliminate the call
to SourceManager::getLocation() while loading a precompiled preamble,
since SourceManager::getLocation() itself causes unwanted
deserialization.
Fixed <rdar://problem/9399352>.
llvm-svn: 131021
- New isDefined() function checks for deletedness
- isThisDeclarationADefinition checks for deletedness
- New doesThisDeclarationHaveABody() does what
isThisDeclarationADefinition() used to do
- The IsDeleted bit is not propagated across redeclarations
- isDeleted() now checks the canoncial declaration
- New isDeletedAsWritten() does what it says on the tin.
- isUserProvided() now correct (thanks Richard!)
This fixes the bug that we weren't catching
void foo() = delete;
void foo() {}
as being a redefinition.
llvm-svn: 131013
CXTranslationUnit_NestedMacroInstantiations, which indicates whether
we want to see "nested" macro instantiations (e.g., those that occur
inside other macro instantiations) within the detailed preprocessing
record. Many clients (e.g., those that only care about visible tokens)
don't care about this information, and in code that uses preprocessor
metaprogramming, this information can have a very high cost.
Addresses <rdar://problem/9389320>.
llvm-svn: 130990
Increase robustness of the delegating constructor cycle detection
mechanism. No more infinite loops on invalid or logic errors leading to
false results. Ensure that this is maintained correctly accross
serialization.
llvm-svn: 130887
which determines whether a particular file is actually a header that
is intended to be guarded from multiple inclusions within the same
translation unit.
llvm-svn: 130808
Decl actually found via name lookup & overload resolution when that Decl
is different from the ValueDecl which is actually referenced by the
expression.
This can be used by AST consumers to correctly attribute references to
the spelling location of a using declaration, and otherwise gain insight
into the name resolution performed by Clang.
The public interface to DRE is kept as narrow as possible: we provide
a getFoundDecl() which always returns a NamedDecl, either the ValueDecl
referenced or the new, more precise NamedDecl if present. This way AST
clients can code against getFoundDecl without know when exactly the AST
has a split representation.
For an example of the data this provides consider:
% cat x.cc
namespace N1 {
struct S {};
void f(const S&);
}
void test(N1::S s) {
f(s);
using N1::f;
f(s);
}
% ./bin/clang -fsyntax-only -Xclang -ast-dump x.cc
[...]
void test(N1::S s) (CompoundStmt 0x5b02010 <x.cc:5:20, line:9:1>
(CallExpr 0x5b01df0 <line:6:3, col:6> 'void'
(ImplicitCastExpr 0x5b01dd8 <col:3> 'void (*)(const struct N1::S &)' <FunctionToPointerDecay>
(DeclRefExpr 0x5b01d80 <col:3> 'void (const struct N1::S &)' lvalue Function 0x5b01a20 'f' 'void (const struct N1::S &)'))
(ImplicitCastExpr 0x5b01e20 <col:5> 'const struct N1::S' lvalue <NoOp>
(DeclRefExpr 0x5b01d58 <col:5> 'N1::S':'struct N1::S' lvalue ParmVar 0x5b01b60 's' 'N1::S':'struct N1::S')))
(DeclStmt 0x5b01ee0 <line:7:3, col:14>
0x5b01e40 "UsingN1::;")
(CallExpr 0x5b01fc8 <line:8:3, col:6> 'void'
(ImplicitCastExpr 0x5b01fb0 <col:3> 'void (*)(const struct N1::S &)' <FunctionToPointerDecay>
(DeclRefExpr 0x5b01f80 <col:3> 'void (const struct N1::S &)' lvalue Function 0x5b01a20 'f' 'void (const struct N1::S &)' (UsingShadow 0x5b01ea0 'f')))
(ImplicitCastExpr 0x5b01ff8 <col:5> 'const struct N1::S' lvalue <NoOp>
(DeclRefExpr 0x5b01f58 <col:5> 'N1::S':'struct N1::S' lvalue ParmVar 0x5b01b60 's' 'N1::S':'struct N1::S'))))
Now we can tell that the second call is 'using' (no pun intended) the using
declaration, and *which* using declaration it sees. Without this, we can
mistake calls that go through using declarations for ADL calls, and have no way
to attribute names looked up with using declarations to the appropriate
UsingDecl.
llvm-svn: 130670
parameter node and use this to correctly mangle parameter
references in function template signatures.
A follow-up patch will improve the storage usage of these
fields; here I've just done the lazy thing.
llvm-svn: 130669
NestedNameSpecifierLoc. It predates when we had such an object.
Reference the NNSLoc directly in DREs, and embed it directly into the
MemberNameQualifier struct.
llvm-svn: 130668
Mostly trailing whitespace so that me editor nuking it doesn't muddy the
waters of subsequent commits that do change functionality.
Also nukes a stray statement that was harmless but redundant that
I introduced in r130666.
llvm-svn: 130667
a bitfield in the base class. DREs weren't using any bits here past the
normal Expr bits, so we have plenty of room. This makes the common case
of getting a Decl out of a DRE no longer need to do any masking etc.
Also, while here, clean up code to use the accessor methods rather than
directly poking these bits, and provide a nice comment for DREs that
includes the information previously attached to the bits going into the
pointer union.
No functionality changed here, but DREs should be a tad faster now.
llvm-svn: 130666
accompanying fixes to make it work today.
The core of this patch is to provide a link from a TemplateTypeParmType
back to the TemplateTypeParmDecl node which declared it. This in turn
provides much more precise information about the type, where it came
from, and how it functions for AST consumers.
To make the patch work almost a year after its first attempt, it needed
serialization support, and it now retains the old getName() interface.
Finally, it requires us to not attempt to instantiate the type in an
unsupported friend decl -- specifically those coming from template
friend decls but which refer to a specific type through a dependent
name.
A cleaner representation of the last item would be to build
FriendTemplateDecl nodes for these, storing their template parameters
etc, and to perform proper instantation of them like any other template
declaration. They can still be flagged as unsupported for the purpose of
access checking, etc.
This passed an asserts-enabled bootstrap for me, and the reduced test
case mentioned in the original review thread no longer causes issues,
likely fixed at somewhere amidst the 24k revisions that have elapsed.
llvm-svn: 130628
type trait. The previous implementation suffered from several problems:
1) It implemented all of the logic in RecordType by walking over every
base and field in a CXXRecordDecl and validating the constraints of
the standard. This made for very straightforward code, but is
extremely inefficient. It also is conceptually wrong, the logic tied
to the C++ definition of standard-layout classes should be in
CXXRecordDecl, not RecordType.
2) To address the performance problems with #1, a cache bit was added to
CXXRecordDecl, and at the completion of every C++ class, the
RecordType was queried to determine if it was a standard layout
class, and that state was cached. Two things went very very wrong
with this. First, the caching version of the query *was never
called*. Even within the recursive steps of the walk over all fields
and bases the caching variant was not called, making each query
a full *recursive* walk. Second, despite the cache not being used, it
was computed for every class declared, even when the trait was never
used in the program. This probably significantly regressed compile
time performance for edge-case files.
3) An ASTContext was required merely to query the type trait because
querying it performed the actual computations.
4) The caching bit wasn't managed correctly (uninitialized).
The new implementation follows the system for all the other traits on
C++ classes by encoding all the state needed in the definition data and
building up the trait incrementally as each base and member are added to
the definition of the class.
The idiosyncracies of the specification of standard-layout classes
requires more state than I would like; currently 5 bits. I could
eliminate one of the bits easily at the expense of both clarity and
resilience of the code. I might be able to eliminate one of the other
bits by computing its state in terms of other state bits in the
definition. I've already done that in one place where there was a fairly
simple way to achieve it.
It's possible some of the bits could be moved out of the definition data
and into some other structure which isn't serialized if the serialized
bloat is a problem. That would preclude serialization of a partial class
declaration, but that's likely already precluded.
Comments on any of these issues welcome.
llvm-svn: 130601
Argyrios Kyrtzidis