Manuel Klimek [Wed, 19 Jun 2013 15:42:45 +0000 (15:42 +0000)]
Completely revamp node binding for AST matchers.
This is in preparation for the backwards references to bound
nodes, which will expose a lot more about how matches occur. Main
changes:
- instead of building the tree of bound nodes, we build a "set" of bound
nodes and explode all possible match combinations while running
through the matchers; this will allow us to also implement matchers
that filter down the current set of matches, like "equalsBoundNode"
- take the set of bound nodes at the start of the match into
consideration when doing memoization; as part of that, reevaluated
that memoization gives us benefits that are large enough (it still
does - the effect on common match patterns is up to an order of
magnitude)
- reset the bound nodes when a node does not match, thus never leaking
information from partial sub-matcher matches for failing matchers
Effects:
- we can now correctly "explode" combinatorial matches, for example:
allOf(forEachDescendant(...bind("a")),
forEachDescendant(...bind("b"))) will now trigger matches for all
combinations of matching "a" and "b"s.
- we now never expose bound nodes from partial matches in matchers that
did not match in the end - this fixes a long-standing issue
FIXMEs:
- rename BoundNodesTreeBuilder to BoundNodesBuilder or
BoundNodesSetBuilder, as we don't build a tree any more; this is out
of scope for this change, though
- we're seeing some performance regressions (around 10%), but I expect
some performance tuning will get that back, and it's easily worth
the increase in expressiveness for now
Reid Kleckner [Wed, 19 Jun 2013 15:20:38 +0000 (15:20 +0000)]
[ms-cxxabi] Emit and install appropriately mangled vbtables
In Itanium, dynamic classes have one vtable with several different
address points for dynamic base classes that can't share vtables.
In the MS C++ ABI, each vbtable that can't be shared gets its own
symbol, similar to how ctor vtables work in Itanium. However, instead
of mangling the subobject offset into the symbol, the unique portions of
the inheritance path are mangled into the symbol to make it unique.
This patch implements the MSVC 2012 scheme for forming unique vbtable
symbol names. MSVC 2010 use the same mangling with a different subset
of the path. Implementing that mangling and possibly others is TODO.
Each vbtable is an array of i32 offsets from the vbptr that points to it
to another virtual base subobject. The first entry of a vbtable always
points to the base of the current subobject, implying that it is the
same no matter which parent class contains it.
Reid Kleckner [Wed, 19 Jun 2013 15:09:06 +0000 (15:09 +0000)]
[Driver] Don't forward source file input args to gcc
gcc's inputs are already added by the InputInfoList passed to
Action::ConstructJob.
Fixes a regression from r183989. This was manifesting when targetting
mingw as an extra input argument to gcc when assembling. It presumably
affects other situations where clang calls gcc.
Prior to r183989, forwardToGCC() was returning false because the INPUT
option defined in OptParser.td had the DriverOption flag set on it.
LLVM's Option library does not set this flag for INPUT.
Manman Ren [Wed, 19 Jun 2013 01:46:49 +0000 (01:46 +0000)]
Debug Info: support for gdwarf-2 gdwarf-3 gdwarf-4
These options will add a module flag with name "Dwarf Version".
The behavior flag is currently set to Warning, so when two values disagree,
a warning will be emitted.
Anna Zaks [Tue, 18 Jun 2013 23:16:15 +0000 (23:16 +0000)]
[analyzer] Do not report uninitialized value warnings inside swap functions.
This silences warnings that could occur when one is swapping partially initialized structs. We suppress
not only the assignments of uninitialized members, but any values inside swap because swap could
potentially be used as a subroutine to swap class members.
This silences a warning from std::try::function::swap() on partially initialized objects.
Eli Friedman [Tue, 18 Jun 2013 22:41:37 +0000 (22:41 +0000)]
Introduce a new mangling for protocol-qualified ObjC types in C++. This allows
to provide proper overloading, and also prevents mangling conflicts with
template arguments of protocol-qualified type.
This is a non-backward-compatible mangling change, but per discussion with
John, the benefits outweigh this cost.
the lookup for 'OldImage' will return the 'NewImage' decl ("@class NewImage").
In such a case, when creating the decl for "@class OldImage" use the real declaration name ("NewImage"),
instead of the alias one ("OldImage"), otherwise we will break IdentifierResolver and redecls-chain invariants.
[multiprecision-builtins] Added missing builtin __builtin_{add,sub}cb for {add,sub} with carry for bytes.
I have had several people ask me about why this builtin was not available in
clang (since it seems like a logical conclusion). This patch implements said
builtins.
Relevant tests are included as well. I also updated the Clang language extension reference.
Richard Smith [Tue, 18 Jun 2013 20:15:12 +0000 (20:15 +0000)]
DR14, DR101, and part of DR1: fix handling of extern "C" declarations in
namespaces, by treating them just like we treat extern "C" declarations in
function scope.
Richard Smith [Tue, 18 Jun 2013 17:51:51 +0000 (17:51 +0000)]
PR14503: Don't assert if a constexpr constructor temploid instantiates to a
constructor that does not initialize all members, and that constructor is used
to initialize a global.
Stephen Lin [Tue, 18 Jun 2013 17:00:49 +0000 (17:00 +0000)]
CodeGen: Have 'this'-returning constructors and destructors to take advantage of the new backend 'returned' attribute.
The backend will now use the generic 'returned' attribute to form tail calls where possible, as well as avoid save-restores of 'this' in some cases (specifically the cases that matter for the ARM C++ ABI).
This patch also reverts a prior front-end only partial implementation of these optimizations, since it's no longer required.
Objective-C [qoi]: privide typo correction for selectors
in addition of receiver having static type, but also when
receiver has dynamic type (of 'id' variety) as well as when
receiver is of 'Class' type vareity. // rdar://7853549
Bob Wilson [Tue, 18 Jun 2013 05:36:04 +0000 (05:36 +0000)]
size_t on Darwin AAPCS targets is "unsigned long". <rdar://problem/14136459>
Some embedded targets use ARM's AAPCS with iOS header files that define size_t
as unsigned long, which conflicts with the usual AAPCS definition of size_t
as unsigned int.
Adrian Prantl [Tue, 18 Jun 2013 00:27:36 +0000 (00:27 +0000)]
Remove an ugly hack that was meant to eliminate the breakpoint ambiguity
between a block assignment and the entry of the block function. In reality
this wouldn't work anyway because blocks are predominantly created
on-the-fly inside of an ObjC method invocation.
The proper fix for the ambiguity is to use -gcolumn-info to differentiate
the breakpoints.
This is expected to break some block-related darwin-gdb tests.
Manman Ren [Mon, 17 Jun 2013 19:42:49 +0000 (19:42 +0000)]
X86 intrinsics: cmpge|gt|nge|ngt_ss|_sd
These intrinsics should return the comparision result in the low bits and keep
the high bits of the first source operand.
When calling to builtin functions, the source operands are swapped and the high
bits of the second source operand are kept. To fix the issue, an extra
shufflevector is used.
Rafael Espindola [Mon, 17 Jun 2013 19:26:19 +0000 (19:26 +0000)]
Revert "Updated test cases for contextual conversion"
This reverts commit r184100.
It was faling on some bots:
http://bb.pgr.jp/builders/cmake-clang-i686-mingw32/builds/1973/steps/test_clang/logs/Clang%20%3A%3A%20SemaCXX__cxx1y-contextual-conversion-tweaks.cpp
Richard Smith [Mon, 17 Jun 2013 18:05:42 +0000 (18:05 +0000)]
Avoid using __SIZE_TYPE__ in a -std=c++98 -pedantic-errors test, because that
triggers an error on MS targets where __SIZE_TYPE__ expands to 'long long'.
Serge Pavlov [Mon, 17 Jun 2013 17:18:51 +0000 (17:18 +0000)]
Clean up empty struct/union recognition.
Make use of getTypeSizeInChars to detect structs/unions of zero size. It allows
more accurate detection of types of zero size. It however has a side effect -
sequence of used types may change, that is why the test 'override-layout' was
modified.
Fixes incorrect indentation of line comments after break and re-alignment.
Summary:
Selectively propagate the information about token kind in
WhitespaceManager::replaceWhitespaceInToken.For correct alignment of new
segments of line comments in order to align them correctly. Don't set
BreakBeforeParameter in breakProtrudingToken for line comments, as it introduces
a break after the _next_ parameter. Added tests for related functions.
Reid Kleckner [Mon, 17 Jun 2013 12:56:08 +0000 (12:56 +0000)]
[AST] Don't include RecursiveASTVisitor.h in ASTContext.h
The untemplated implementation of getParents() doesn't need to be in a
header file.
RecursiveASTVisitor.h is full of repeated macro expansion. Moving this
include to ASTContext.cpp speeds up compilation of
LambdaMangleContext.cpp, a small C++ file with few includes, from 3.7s
to 2.8s for me locally. I haven't measured a full build, but it can't
hurt.
I had to fix a few static analyzer files that were depending on
transitive includes of C++ AST headers.
Richard Smith [Sun, 16 Jun 2013 05:05:39 +0000 (05:05 +0000)]
PR16339: Don't produce a diagnostic pointing at the whitespace between a '#if'
and a '!defined(X)' if we find a broken header guard. This is suboptimal; we
should point the diagnostic at the 'X' token not the 'if' token, but it fixes
the crash.
Larisse Voufo [Sat, 15 Jun 2013 20:17:46 +0000 (20:17 +0000)]
Updated the support for contextual conversion tweaks (n3323) with a previously overlooked part: implicitly converting array sizes to size_t, rather than contextually converting them to some unique type.
Faisal Vali [Sat, 15 Jun 2013 11:54:37 +0000 (11:54 +0000)]
A quick fix to allow return type deduction on member templates
by ensuring DiagnoseUseOfDecl is called both on the found decl and the
decl being used (i.e the specialization in the case of member templates) whenever they are different.
Per the exchange captured in
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20130610/081636.html
a more comprehensive fix that allows both decls to be passed into DiagnoseUseOfDecl is (or should be) forthcoming relatively soon.
Ed Schouten [Sat, 15 Jun 2013 09:40:14 +0000 (09:40 +0000)]
Emit native implementations of atomic operations on FreeBSD/armv6.
Just like on Linux, FreeBSD/armv6 assumes the system supports
ldrex/strex unconditionally. It is also used by the kernel. We can
therefore enable support for it, like we do on Linux.
While there, change one of the unit tests to explicitly test against
armv5 instead of armv7, as it actually tests whether libcalls are
emitted.
Richard Smith [Sat, 15 Jun 2013 00:30:29 +0000 (00:30 +0000)]
PR16263: Implement current direction of core issue 1376. Binding a reference to
the result of a cast-to-reference-type lifetime-extends the object to which the
reference inside the cast binds.
This requires us to look for subobject adjustments on both the inside and the
outside of the MaterializeTemporaryExpr when looking for a temporary to
lifetime-extend (which we also need for core issue 616, and possibly 1213).
Richard Smith [Fri, 14 Jun 2013 22:27:52 +0000 (22:27 +0000)]
Fix handling of const_cast from prvalue to rvalue reference: such a cast is
only permitted if the source object is of class type, and should materialize a
temporary for the reference to bind to.
Reid Kleckner [Fri, 14 Jun 2013 17:17:23 +0000 (17:17 +0000)]
[Driver] Refactor clang driver to use LLVM's Option library
The big changes are:
- Deleting Driver/(Arg|Opt)*
- Rewriting includes to llvm/Option/ and re-sorting
- 'using namespace llvm::opt' in clang::driver
- Fixing the autoconf build by adding option everywhere
As discussed in the review, this change includes using directives in
header files. I'll make follow up changes to remove those in favor of
name specifiers.
Summary:
Don't remove backslashes from block comments. Previously this
/* \ \ \ \ \ \
*/
would be turned to this:
/*
*/
which spoils some kinds of ASCII-art, people use in their comments. The behavior
was related to handling escaped newlines in block comments inside preprocessor
directives. This patch makes handling it in a more civilized way.
Chandler Carruth [Fri, 14 Jun 2013 08:57:18 +0000 (08:57 +0000)]
Fix the warning for divide by zero to be a bit more robust. ;]
Previously, it only ever fired for zeros which formed null pointers.
Now, hilariously, in C++98 this was almost anything. Including tricks
like warning on the divisor in this code:
Why the RHS of the outer divide is a null pointer constant is a sordid
tale of sorrow. Anyways, the committee fixed this for C++11 and onward
as part of core isssue 903, and Richard recently implemented this fix
causing the warning to go away here (and elsewhere).
This patch restores the warning here and adds it for numerous other
somewhat obvious gaffes:
int g(int x) {
return x / (int)(0.0);
}
The patch is essentially just using the full power of our constant
folding in Clang to produce the warning, but insisting that it must fold
to an *integer* which is zero so that we don't get false positives
anywhere.
Richard Smith [Thu, 13 Jun 2013 22:07:02 +0000 (22:07 +0000)]
Avoid fallthrough in header, to allow external users of Clang libraries to
build with Clang's -Wimplicit-fallthrough warning enabled. The fallthrough
was not making this code better.
projects / clang / log