Daniel Dunbar [Sat, 7 Feb 2009 02:46:03 +0000 (02:46 +0000)]
Set load/store alignment when doing ABI coercions.
- Currently, this is producing poor code, but we prefer correctness
to performance for now. Eventually we should be able to generally
avoid having to set the alignment when we control the alignment of
the alloca.
- This knocks out 33/1000 failures on my single argument ABI tests,
down to 22/1000 and 18 of these appear to be gcc bugs. Woot.
Ted Kremenek [Sat, 7 Feb 2009 01:47:29 +0000 (01:47 +0000)]
Overhaul of Stmt allocation:
- Made allocation of Stmt objects using vanilla new/delete a *compiler
error* by making this new/delete "protected" within class Stmt.
- Now the only way to allocate Stmt objects is by using the new
operator that takes ASTContext& as an argument. This ensures that
all Stmt nodes are allocated from the same (pool) allocator.
- Naturally, these two changes required that *all* creation sites for
AST nodes use new (ASTContext&). This is a large patch, but the
majority of the changes are just this mechanical adjustment.
- The above changes also mean that AST nodes can no longer be
deallocated using 'delete'. Instead, one most do
StmtObject->Destroy(ASTContext&) or do
ASTContextObject.Deallocate(StmtObject) (the latter not running the
'Destroy' method).
Along the way I also...
- Made CompoundStmt allocate its array of Stmt* using the allocator in
ASTContext (previously it used std::vector). There are a whole
bunch of other Stmt classes that need to be similarly changed to
ensure that all memory allocated for ASTs comes from the allocator
in ASTContext.
- Added a new smart pointer ExprOwningPtr to Sema.h. This replaces
the uses of llvm::OwningPtr within Sema, as llvm::OwningPtr used
'delete' to free memory instead of a Stmt's 'Destroy' method.
Big thanks to Doug Gregor for helping with the acrobatics of making
'new/delete' private and the new smart pointer ExprOwningPtr!
Ted Kremenek [Sat, 7 Feb 2009 00:52:24 +0000 (00:52 +0000)]
GRExprEngine: When processing compound assignments, do a switch table lookup to get the non-compound opcode from the compound opcode instead of relying on the order of BinaryOperator::opcode values. This unbreaks the misc-ps.c test.
Chris Lattner [Sat, 7 Feb 2009 00:23:17 +0000 (00:23 +0000)]
Apparently it is important to define intptr_t and uintptr_t to
long instead of int. This is because system heaers like to redefine
typedefs and that is an error if they don't exactly match. Use long
for intptr_t on all systems where long is the right size.
Chris Lattner [Fri, 6 Feb 2009 23:29:39 +0000 (23:29 +0000)]
only define MB_LEN_MAX if the system <limits.h> doesn't.
don't typecast CHAR_MIN to char, this makes it not a PP constant
and gives it the wrong unpromoted type. Thanks to Sebastian for
pointing this out!
Chris Lattner [Fri, 6 Feb 2009 22:59:26 +0000 (22:59 +0000)]
Export __INT8_TYPE__ / __INT16_TYPE__ / __INT32_TYPE__ / __INT64_TYPE__
in a gcc 4.5 compatible way so that stdint.h can follow the compiler's
notion of types.
Douglas Gregor [Fri, 6 Feb 2009 22:42:48 +0000 (22:42 +0000)]
Semantic checking for class template declarations and
redeclarations. For example, checks that a class template
redeclaration has the same template parameters as previous
declarations.
Detangled class-template checking from ActOnTag, whose logic was
getting rather convoluted because it tried to handle C, C++, and C++
template semantics in one shot.
Made some inroads toward eliminating extraneous "declaration does not
declare anything" errors by adding an "error" type specifier.
Ted Kremenek [Fri, 6 Feb 2009 19:55:15 +0000 (19:55 +0000)]
Move StringLiteral to allocate its internal string data using the allocator in
ASTContext. This required changing all clients to pass in the ASTContext& to the
constructor of StringLiteral. I also changed all allocations of StringLiteral to
use new(ASTContext&).
Along the way, I updated a bunch of new()'s in StmtSerialization.cpp to use the
allocator from ASTContext& (not complete).
Douglas Gregor [Thu, 5 Feb 2009 23:33:38 +0000 (23:33 +0000)]
Improve the representation of template type parameters. We now
canonicalize by template parameter depth, index, and name, and the
unnamed version of a template parameter serves as the canonical.
TemplateTypeParmDecl no longer needs to inherit from
TemplateParmPosition, since depth and index information is present
within the type.
Sebastian Redl [Thu, 5 Feb 2009 15:12:41 +0000 (15:12 +0000)]
Fix the symptom of the regression, by having the CXXConditionDeclExpr not destroy its Decl.
However, the cause still remains: the Decl is linked into the chain of its DeclContext and remains there despite being deleted.
Sebastian Redl [Thu, 5 Feb 2009 15:02:23 +0000 (15:02 +0000)]
Put the invalid flag of OwningResult into the Action pointer.
This shrinks OwningResult by one pointer. Since it is no longer larger than OwningPtr, merge the two.
This leads to simpler client code and speeds up my benchmark by 2.7%.
For some reason, this exposes a previously hidden bug, causing a regression in SemaCXX/condition.cpp.
Daniel Dunbar [Thu, 5 Feb 2009 11:21:33 +0000 (11:21 +0000)]
Add bare bones test that parameter passing is consistent for
scalar/complex/aggregate cases.
- Currently disabled for x86_64, triggering a misoptimization
(PR3489).
Daniel Dunbar [Thu, 5 Feb 2009 11:13:54 +0000 (11:13 +0000)]
Implement Direct ABIInfo semantics.
- No intended functionality change, this is essentially enabling
direct passing of complex and aggregate values, which no ABI is
using.
Daniel Dunbar [Thu, 5 Feb 2009 09:16:39 +0000 (09:16 +0000)]
Implement ABI Indirect sematics for arguments.
- No intended functionality change, all current ABI implementations
were only using indirect for complex/aggregate types, which were
being passed indirectly with the Direct ABIInfo kind.
Zhongxing Xu [Thu, 5 Feb 2009 06:57:29 +0000 (06:57 +0000)]
Make SymbolicRegion subclass TypedRegion, for symbols usually have types, so
do the symblic regions associated with them and we need them to be typed.
Current SymbolicRegion::getRValueType() method is very restricting. It may be
modified when we are more clear about what could be the types of symblic
regions.
BasicConstraintManager::Assume() is changed due to that now SymblicRegion is a
subclass of SubRegion.
Ted Kremenek [Thu, 5 Feb 2009 06:50:21 +0000 (06:50 +0000)]
Remove a bunch of obscene double-buffering of BugReports in the retain/release
checker. This was previously needed because BugReport objects were previously
allocated on the stack and not owned by BugReporter. Now we can just issue them
on the fly. This change was motivated because we were seeing some weird cases
where some really long paths would get issued for bugs (particularly leaks)
because of some double-caching.
Ted Kremenek [Wed, 4 Feb 2009 23:49:09 +0000 (23:49 +0000)]
Overhaul BugReporter interface and implementation. The new interface cleans up
the ownership of BugTypes and BugReports. Now BugReports are owned by BugTypes,
and BugTypes are owned by the BugReporter object.
The major functionality change in this patch is that reports are not immediately
emitted by a call to BugReporter::EmitWarning (now called EmitReport), but
instead of queued up in report "equivalence classes". When
BugReporter::FlushReports() is called, it emits one diagnostic per report
equivalence class. This provides a nice cleanup with the caching of reports as
well as enables the BugReporter engine to select the "best" path for reporting a
path-sensitive bug based on all the locations in the ExplodedGraph that the same
bug could occur.
Along with this patch, Leaks are now coalesced into a common equivalence class
by their allocation site, and the "summary" diagnostic for leaks now reports the
allocation site as the location of the bug (this may later be augmented to also
provide an example location where the leak occurs).
Douglas Gregor [Wed, 4 Feb 2009 22:46:25 +0000 (22:46 +0000)]
Implement semantic analysis for the GNU flexible array initialization
extension. The interaction with designated initializers is a
bit... interesting... but we follow GNU's lead and don't permit too
much crazy code in this area.
Also, make the "excess initializers" error message a bit more
informative.
Daniel Dunbar [Wed, 4 Feb 2009 21:19:06 +0000 (21:19 +0000)]
Add -femit-all-decls codegen option.
- Emits all declarations, even unused (static) ones.
- Useful when doing minimization of codegen problems (otherwise
problems localized to a static function aren't minimized well).
projects / clang / log