Fix the memory leak of FloatingLiteral/IntegerLiteral.
For large floats/integers, APFloat/APInt will allocate memory from the heap to represent these numbers.
Unfortunately, when we use a BumpPtrAllocator to allocate IntegerLiteral/FloatingLiteral nodes the memory associated with
the APFloat/APInt values will never get freed.
I introduce the class 'APNumericStorage' which uses ASTContext's allocator for memory allocation and is used internally by FloatingLiteral/IntegerLiteral.
John McCall [Sat, 28 Aug 2010 07:56:00 +0000 (07:56 +0000)]
When checking access control for an instance member access on
an object of type I, if the current access target is protected
when named in a class N, consider the friends of the classes P
where I <= P <= N and where a notional member of N would be
non-forbidden in P.
Ted Kremenek [Sat, 28 Aug 2010 00:19:02 +0000 (00:19 +0000)]
Explicitly handle CXXExprWithTemporaries during CFG construction by just visiting the subexpression. While we don't do anything intelligent right now, this obviates a bogus -Wunreahable-code warning reported in PR 6130.
Tom Care [Fri, 27 Aug 2010 22:46:32 +0000 (22:46 +0000)]
Enabled relaxed LiveVariables analysis in the path-sensitive engine to increase the coverage of bugs. Primarily affects IdempotentOperationChecker.
- Migrated a temporarily separated test back to its original file (bug has been fixed, null-deref-ps-temp.c -> null-deref-ps.c)
- Changed SymbolManager to use relaxed LiveVariables
- Updated several test cases that the IdempotentOperationChecker class now flags
- Added test case to test relaxed LiveVariables use by the IdempotentOperationChecker
Tom Care [Fri, 27 Aug 2010 22:30:10 +0000 (22:30 +0000)]
Add alternate version of LiveVariables analysis that does not kill liveness at assignments. This 'relaxed' liveness is useful in path sensitive analysis for situations where the resulting extended liveness allows us to find some bugs.
- Added killAtAssign flag to LiveVariables
- Added relaxed LiveVariables to AnalysisContext with an accessor
Ted Kremenek [Fri, 27 Aug 2010 21:34:58 +0000 (21:34 +0000)]
Implement CXCursor support for walking C++ base specifiers. This includes adding the API hooks clang_isVirtualBase() and clang_getCXXAccessSpecifier() to query properties of the base specifier.
Douglas Gregor [Fri, 27 Aug 2010 21:18:54 +0000 (21:18 +0000)]
Implement the "call super" code completion for C++. If the virtual
member function you're typing in overrides another virtual function,
this fills in a (qualified!) call to that virtual function to make
such delegation easy.
Douglas Gregor [Fri, 27 Aug 2010 15:10:57 +0000 (15:10 +0000)]
Add a super-cool code completion for send-to-super. When we're typing
a message send to "super" from a method that appears to be meant to
override a superclass method (same kind, same selector, same argument
types), provide a "super" completion that fills in the selector along
with forwarding the method's arguments (as placeholders).
John McCall [Fri, 27 Aug 2010 09:08:28 +0000 (09:08 +0000)]
Propagate whether an id-expression is the immediate argument of
an '&' expression from the second caller of ActOnIdExpression.
Teach template argument deduction that an overloaded id-expression
doesn't give a valid type for deduction purposes to a non-static
member function unless the expression has the correct syntactic
form.
Teach ActOnIdExpression that it shouldn't try to create implicit
member expressions for '&function', because this isn't a
permitted form of use for member functions.
Teach CheckAddressOfOperand to diagnose these more carefully.
Some of these cases aren't reachable right now because earlier
diagnostics interrupt them.
John McCall [Thu, 26 Aug 2010 23:41:50 +0000 (23:41 +0000)]
One who seeks knowledge learns something new every day.
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.
Ted Kremenek [Thu, 26 Aug 2010 22:19:33 +0000 (22:19 +0000)]
Fix horrible GRExprEngine bug where switch statements with no 'case:' statements would cause the path to get prematurely aborted. Fixes <rdar://problem/8360854>.
Chris Lattner [Thu, 26 Aug 2010 20:05:13 +0000 (20:05 +0000)]
fix 2xi16 to pass as i32 instead of <2 x i16>. The former passes in
memory (as required) the later now passes in an xmm register. This
fixes gcc.dg/compat/vector_1 on x86-32.
John McCall [Thu, 26 Aug 2010 17:42:30 +0000 (17:42 +0000)]
...I forgot to check my new test after adding it, and lo, there's slightly different
behavior in C than in C++ (which is what the original test case was).
Douglas Gregor [Thu, 26 Aug 2010 16:36:48 +0000 (16:36 +0000)]
When code-completing a potential call to a C++ non-static member
function, take into account the qualifiers on the object argument
(e.g., what will become "this"), filtering around uncallable member
functions and giving a slight priority boost to those with
exactly-matching qualifiers.
Douglas Gregor [Thu, 26 Aug 2010 13:48:20 +0000 (13:48 +0000)]
Move the sorting of code-completion results out of the main path and
into the clients, e.g., the printing code-completion consumer and
c-index-test. Clients may want to re-sort the results anyway.
Provide a libclang function that sorts the results.
Douglas Gregor [Thu, 26 Aug 2010 02:23:45 +0000 (02:23 +0000)]
Move the sorting of code-completion results out of the main path and
into the clients, e.g., the printing code-completion consumer and
c-index-test. Clients may want to re-sort the results anyway.
Provide a libclang function that sorts the results.
Eric Christopher [Thu, 26 Aug 2010 00:42:16 +0000 (00:42 +0000)]
With lax vector conversions (the default) make sure we convert between two
vectors that are the same size. Fix up testcases accordingly and add a new one
to make sure we still error if lax vector conversions are disabled.
Douglas Gregor [Thu, 26 Aug 2010 00:30:24 +0000 (00:30 +0000)]
Revert "Move the sorting of code-completion results out of the main path and
into the clients", because the C standard library sucks. Where's my
stable sort, huh?
John McCall [Wed, 25 Aug 2010 23:44:00 +0000 (23:44 +0000)]
Pull DelayedDiagnostic and AccessedEntity out into their own header.
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.
Chris Lattner [Wed, 25 Aug 2010 23:39:14 +0000 (23:39 +0000)]
Finally pass "two floats in a 64-bit unit" as a <2 x float> instead of
as a double in the x86-64 ABI. This allows us to generate much better
code for certain things, e.g.:
_Complex float f32(_Complex float A, _Complex float B) {
return A+B;
}
Used to compile into (look at the integer silliness!):
There is still "stuff to be done" here for the struct case,
but this resolves rdar://6379669 - [x86-64 ABI] Pass and return
_Complex float / double efficiently
Tom Care [Wed, 25 Aug 2010 22:37:26 +0000 (22:37 +0000)]
Improved the handling of blocks and block variables in PseudoConstantAnalysis
- Removed the assumption that __block vars are all non-constant
- Simplified some repetitive code in RunAnalysis
- Added block walking support
- Code/comments cleanup
- Separated out test for block pseudoconstants
projects / clang / log