access control errors into SFINAE errors, so that the trait provides
enough support to implement the C++0x std::is_convertible type trait.
To get there, the SFINAETrap now knows how to set up a SFINAE context
independent of any template instantiations or template argument
deduction steps, and (separately) can set a Sema flag to translate
access control errors into SFINAE errors. The latter can also be
useful if we decide that access control errors during template argument
deduction should cause substitution failure (rather than a hard error)
as has been proposed for C++0x.
llvm-svn: 124446
whose patterns are template arguments. We can now instantiate, e.g.,
typedef tuple<pair<OuterTypes, InnerTypes>...> type;
where OuterTypes and InnerTypes are template type parameter packs.
There is a horrible inefficiency in
TemplateArgumentLoc::getPackExpansionPattern(), where we need to
create copies of TypeLoc data because our interfaces traffic in
TypeSourceInfo pointers where they should traffic in TypeLocs
instead. I've isolated in efficiency in this one routine; once we
refactor our interfaces to traffic in TypeLocs, we can eliminate it.
llvm-svn: 122278
a translation unit to the ActOnEndOfTranslationUnit function instead of doing
it at the start of DefineUsedVTables. The latter is now called *recursively*
during template instantiation, which causes an absolutely insane number of
walks of every record decl in the translation unit.
After this patch, an extremely template instantiation heavy test case's compile
time drops by 10x, and we see between 15% and 20% improvement in average
compile times across a project. This is just recovering a regression, it
doesn't make anything faster than it was several weeks ago.
llvm-svn: 121644
a useful template instantiation stack. Fixes PR8640.
This also causes a slight change to where the "instantianted from" note shows up
in truly esoteric cases (see the change to test/SemaCXX/destructor.cpp), but
that isn't directly the fault of this patch.
llvm-svn: 120135
-Move the stuff of Diagnostic related to creating/querying diagnostic IDs into a new DiagnosticIDs class.
-DiagnosticIDs can be shared among multiple Diagnostics for multiple translation units.
-The rest of the state in Diagnostic object is considered related and tied to one translation unit.
-Have Diagnostic point to the SourceManager that is related with. Diagnostic can now accept just a
SourceLocation instead of a FullSourceLoc.
-Reflect the changes to various interfaces.
llvm-svn: 119730
that are suppressed during template argument deduction. This change
queues diagnostics computed during template argument deduction. Then,
if the resulting function template specialization or partial
specialization is chosen by overload resolution or partial ordering
(respectively), we will emit the queued diagnostics at that point.
This addresses most of PR6784. However, the check for unnamed/local
template arguments (which existed before this change) is still only
skin-deep, and needs to be extended to look deeper into types. It must
be improved to finish PR6784.
llvm-svn: 116373
typeid expressions:
- make sure we have a proper source location for the closing ')'
- cache the declaration of std::type_info once we've found it
llvm-svn: 113441
One who seeks the Tao unlearns something new every day.
Less and less remains until you arrive at non-action.
When you arrive at non-action,
nothing will be left undone.
llvm-svn: 112244
This works courtesy of the new SmallVector<..., 0> specialization that
doesn't require a complete type. Note that you'll need to pull at least
SmallVector.h from LLVM to compile successfully.
llvm-svn: 112114
token. The first token might be something that ends up triggering code
completion, which in turn requires a valid Scope. Test case forthcoming.
llvm-svn: 112066
This option is not part of the Unused diagnostic group until the warnings on llvm codebase are fixed
and we are ready to turn it on. Suggestion by Daniel.
llvm-svn: 111298
Unused warnings for functions:
-static functions
-functions in anonymous namespace
-class methods in anonymous namespace
-class method specializations in anonymous namespace
-function specializations in anonymous namespace
Unused warnings for variables:
-static variables
-variables in anonymous namespace
-static data members in anonymous namespace
-static data members specializations in anonymous namespace
Reveals lots of opportunities for dead code removal in llvm codebase that will
interest my esteemed colleagues.
llvm-svn: 111086
when the CXTranslationUnit_CacheCompletionResults option is given to
clang_parseTranslationUnit(). Essentially, we compute code-completion
results for macro definitions after we have parsed the file, then
store an ASTContext-agnostic version of those results (completion
string, cursor kind, priority, and active contexts) in the
ASTUnit. When performing code completion in that ASTUnit, we splice
the macro definition results into the results provided by the actual
code-completion (which has had macros turned off) before libclang gets
those results. We use completion context information to only splice in
those results that make sense for that context.
With a completion involving all of the macros from Cocoa.h and a few other
system libraries (totally ~8500 macro definitions) living in a
precompiled header, we get about a 9% performance improvement from
code completion, since we no longer have to deserialize all of the
macro definitions from the precompiled header.
Note that macro definitions are merely the canary; the cache is
designed to also support other top-level declarations, which should be
a bigger performance win. That optimization will be next.
Note also that there is no mechanism for determining when to throw
away the cache and recompute its contents.
llvm-svn: 111051
-static variables
-variables in anonymous namespace (fixes rdar://7794535)
-static data members in anonymous namespace
-static data members specializations in anonymous namespace
llvm-svn: 111027
-static function declarations
-functions in anonymous namespace
-class methods in anonymous namespace
-class method specializations in anonymous namespace
-function specializations in anonymous namespace
llvm-svn: 111026
used when parsing (or re-parsing) a file. Also, when loading a
precompiled header into ASTUnit, create a Sema object that holds onto
semantic-analysis information.
llvm-svn: 111003
can create (and hold on to) the Sema object. Also, move Sema-related
initialization/finalization with its various consumers and external
sources into the Sema constructor and destructor, rather than placing
it in ParseAST.
llvm-svn: 110973
can create (and hold on to) the Sema object. Also, move Sema-related
initialization/finalization with its various consumers and external
sources into the Sema constructor and destructor, rather than placing
it in ParseAST.
llvm-svn: 110952
and create separate decl nodes for forward declarations and the
definition," which appears to be causing significant Objective-C
breakage.
llvm-svn: 110803
- Eagerly create ObjCInterfaceTypes for declarations.
- The two above changes lead to a 0.5% increase in memory use and no speed regression when parsing Cocoa.h. On the other hand, now chained PCH works when there's a forward declaration in one PCH and the interface definition in another.
- Add HandleInterestingDecl to ASTConsumer. PCHReader passes the "interesting" decls it finds to this function instead of HandleTopLevelDecl. The default implementation forwards to HandleTopLevelDecl, but ASTUnit's handler for example ignores them. This fixes a potential crash when lazy loading of PCH data would cause ASTUnit's "top level" declaration collection to change while being iterated.
llvm-svn: 110610
This takes some trickery since CastExpr has subclasses (and indeed,
is abstract).
Also, smoosh the CastKind into the bitfield from Expr.
Drops two words of storage from Expr in the common case of expressions
which don't need inheritance paths. Avoids a separate allocation and
another word of overhead in cases needing inheritance paths. Also has
the advantage of not leaking memory, since destructors for AST nodes are
never run.
llvm-svn: 110507
a switch or goto somewhere in the function. Indirect gotos trigger the
jump-checker regardless, because the conditions there are slightly more
elaborate and it's too marginal a case to be worth optimizing.
Turns off the jump-checker in a lot of cases in C++. rdar://problem/7702918
llvm-svn: 109962
Currently, adding it to visible decls of a PCH'ed translation unit has no effect because
adding visible decls before deserialization has no effect (the decls won't be visible).
This will be fixed in a future commit; then it will force deserialization of visible decls, so avoid pointlessly installing it.
llvm-svn: 107595
allows Sema some limited access to the current scope, which we only
use in one way: when Sema is performing some kind of declaration that
is not directly driven by the parser (e.g., due to template
instantiatio or lazy declaration of a member), we can find the Scope
associated with a DeclContext, if that DeclContext is still in the
process of being parsed.
Use this to make the implicit declaration of special member functions
in a C++ class more "scope-less", rather than using the NULL Scope hack.
llvm-svn: 107491
This is more targeted, as it simply provides toggle actions for the parser to
turn access checking on and off. We then use these to suppress access checking
only while we parse the template-id (included scope specifier) of an explicit
instantiation and explicit specialization of a class template. The
specialization behavior is an extension, as it seems likely a defect that the
standard did not exempt them as it does explicit instantiations.
This allows the very common practice of specializing trait classes to work for
private, internal types. This doesn't address instantiating or specializing
function templates, although those apparently already partially work.
The naming and style for the Action layer isn't my favorite, comments and
suggestions would be appreciated there.
llvm-svn: 106993
attribute as part of the calculation. Sema::MarkDeclReferenced(), and
a few other places, want only to consider the "used" bit to determine,
e.g, whether to perform template instantiation. Fixes a linkage issue
with Boost.Serialization.
llvm-svn: 106252
bring in the entire lookup table at once.
Also, give ExternalSemaSource's vtable a home. This is important because otherwise
any reference to it will cause RTTI to be emitted, and since clang is compiled
with -fno-rtti, that RTTI will contain unresolved references (to ExternalASTSource's
RTTI). So this change makes it possible to subclass ExternalSemaSource from projects
compiled with RTTI, as long as the subclass's home is compiled with -fno-rtti.
llvm-svn: 105268
ObjCObjectType, which is basically just a pair of
one of {primitive-id, primitive-Class, user-defined @class}
with
a list of protocols.
An ObjCObjectPointerType is therefore just a pointer which always points to
one of these types (possibly sugared). ObjCInterfaceType is now just a kind
of ObjCObjectType which happens to not carry any protocols.
Alter a rather large number of use sites to use ObjCObjectType instead of
ObjCInterfaceType. Store an ObjCInterfaceType as a pointer on the decl rather
than hashing them in a FoldingSet. Remove some number of methods that are no
longer used, at least after this patch.
By simplifying ObjCObjectPointerType, we are now able to easily remove and apply
pointers to Objective-C types, which is crucial for a certain kind of ObjC++
metaprogramming common in WebKit.
llvm-svn: 103870
return value optimization. Sema marks return statements with their
NRVO candidates (which may or may not end up using the NRVO), then, at
the end of a function body, computes and marks those variables that
can be allocated into the return slot.
I've checked this locally with some debugging statements (not
committed), but there won't be any tests until CodeGen comes along.
llvm-svn: 103865
"used" (e.g., we will refer to the vtable in the generated code) and
when they are defined (i.e., because we've seen the key function
definition). Previously, we were effectively tracking "potential
definitions" rather than uses, so we were a bit too eager about emitting
vtables for classes without key functions.
The new scheme:
- For every use of a vtable, Sema calls MarkVTableUsed() to indicate
the use. For example, this occurs when calling a virtual member
function of the class, defining a constructor of that class type,
dynamic_cast'ing from that type to a derived class, casting
to/through a virtual base class, etc.
- For every definition of a vtable, Sema calls MarkVTableUsed() to
indicate the definition. This happens at the end of the translation
unit for classes whose key function has been defined (so we can
delay computation of the key function; see PR6564), and will also
occur with explicit template instantiation definitions.
- For every vtable defined/used, we mark all of the virtual member
functions of that vtable as defined/used, unless we know that the key
function is in another translation unit. This instantiates virtual
member functions when needed.
- At the end of the translation unit, Sema tells CodeGen (via the
ASTConsumer) which vtables must be defined (CodeGen will define
them) and which may be used (for which CodeGen will define the
vtables lazily).
From a language perspective, both the old and the new schemes are
permissible: we're allowed to instantiate virtual member functions
whenever we want per the standard. However, all other C++ compilers
were more lazy than we were, and our eagerness was both a performance
issue (we instantiated too much) and a portability problem (we broke
Boost test cases, which now pass).
Notes:
(1) There's a ton of churn in the tests, because the order in which
vtables get emitted to IR has changed. I've tried to isolate some of
the larger tests from these issues.
(2) Some diagnostics related to
implicitly-instantiated/implicitly-defined virtual member functions
have moved to the point of first use/definition. It's better this
way.
(3) I could use a review of the places where we MarkVTableUsed, to
see if I missed any place where the language effectively requires a
vtable.
Fixes PR7114 and PR6564.
llvm-svn: 103718
"bottom-up" when implicit casts and comparisons are inserted, compute them
"top-down" when the full expression is finished. Makes it easier to
coordinate warnings and thus implement -Wconversion for signedness
conversions without double-warning with -Wsign-compare. Also makes it possible
to realize that a signedness conversion is okay because the context is
performing the inverse conversion. Also simplifies some logic that was
trying to calculate the ultimate comparison/result type and getting it wrong.
Also fixes a problem with the C++ explicit casts which are often "implemented"
in the AST with a series of implicit cast expressions.
llvm-svn: 103174
the implicit template instantiations we need to perform. Otherwise, we
end up erroneously diagnosing static functions as used if they were
only used within an implicit template instantiation. Fixes a bunch of
spurious failures when building Clang with Clang.
llvm-svn: 100872
how to handle a diagnostic during template argument deduction, which
may be "substitution failure", "suppress", or "report". This keeps us
from, e.g., emitting warnings while performing template argument
deduction.
llvm-svn: 99560
ranges as part of the ASTContext. This code is not and was never used,
but contributes ~250k to the size of the Cocoa.h precompiled
header.
llvm-svn: 99007
which has the label map, switch statement stack, etc. Previously, we
had a single set of maps in Sema (for the function) along with a stack
of block scopes. However, this lead to funky behavior with nested
functions, e.g., in the member functions of local classes.
The explicit-stack approach is far cleaner, and we retain a 1-element
cache so that we're not malloc/free'ing every time we enter a
function. Fixes PR6382.
Also, tweaked the unused-variable warning suppression logic to look at
errors within a given Scope rather than within a given function. The
prior code wasn't looking at the right number-of-errors count when
dealing with blocks, since the block's count would be deallocated
before we got to ActOnPopScope. This approach works with nested
blocks/functions, and gives tighter error recovery.
llvm-svn: 97518
a fixme and PR6451.
Only perform jump checking if the containing function has no errors,
and add the infrastructure needed to do this.
On the testcase in the PR, we produce:
t.cc:6:3: error: illegal goto into protected scope
goto later;
^
t.cc:7:5: note: jump bypasses variable initialization
X x;
^
llvm-svn: 97497
This is to address a serious performance problem observed when running
'clang -fsyntax-only' on really broken source files. In one case,
repeatedly calling CorrectTypo() caused one source file to be rejected
after 2 minutes instead of 1 second.
This patch causes typo correction to take neglible time on that file
while still providing correction results for the first 20 cases. I
felt this was a reasonable number for moderately broken source files.
I don't claim this is the best solution. Comments welcome. It is
necessary for us to address this issue because it is a serious
performance problem.
llvm-svn: 95049
1. Add helper class for sema checks for target attributes
2. Add helper class for codegen of target attributes
As a proof-of-concept - implement msp430's 'interrupt' attribute.
llvm-svn: 93118
maintains a stack of evaluation contexts rather than having the parser
do it. This change made it simpler to track in which contexts
temporaries were created, so that we could...
"Forget" about temporaries created within unevaluated contexts, so
that we don't build a CXXExprWithTemporaries and, therefore, destroy
the integral-constness of our expressions. Fixes PR5609.
llvm-svn: 89908
type and fixes a long-standing code gen. crash reported in
at least two PRs and a radar. (radar 7405040 and pr5025).
There are couple of remaining issues that I would like for
Ted. and Doug to look at:
Ted, please look at failure in Analysis/MissingDealloc.m.
I have temporarily added an expected-warning to make the
test pass. This tests has a declaration of 'SEL' type which
may not co-exist with the new changes.
Doug, please look at a FIXME in PCHWriter.cpp/PCHReader.cpp.
I think the changes which I have ifdef'ed out are correct. They
need be considered for in a few Indexer/PCH test cases.
llvm-svn: 89561
- Provide Sema in callbacks, instead of requiring it in constructor. This
eliminates the need for a factory function. Clients now just pass the object
to consume the results in directly.
- CodeCompleteConsumer is cheap to construct, so building it whenever we are
doing code completion is reasonable.
Doug, please review.
llvm-svn: 87099
with its corresponding template parameter. This can happen when we
performed some substitution into the default template argument and
what we had doesn't match any more, e.g.,
template<int> struct A;
template<typename T, template<T> class X = A> class B;
B<long> b;
Previously, we'd emit a pretty but disembodied diagnostic showing how
the default argument didn't match the template parameter. The
diagnostic was good, but nothing tied it to the *use* of the default
argument in "B<long>". This commit fixes that.
Also, tweak the counting of active template instantiations to avoid
counting non-instantiation records, such as those we create for
(surprise!) checking default arguments, instantiating default
arguments, and performing substitutions as part of template argument
deduction.
llvm-svn: 86884
Peter Collingbourne