Chris Lattner [Sun, 28 Feb 2010 18:18:36 +0000 (18:18 +0000)]
Implement PR6423 by using one token of lookahead to disambiguate
an *almost* always incorrect case. This only does the lookahead
in the insanely unlikely case, so it shouldn't impact performance.
On this testcase:
struct foo {
}
typedef int x;
Before:
t.c:3:9: error: cannot combine with previous 'struct' declaration specifier
typedef int x;
^
John McCall [Sun, 28 Feb 2010 13:00:19 +0000 (13:00 +0000)]
Support constant-evaluation of __builtin_nans* as well as the correct constant
evaluation of __builtin_nan*. Most of the work to make this work is in LLVM.
Fixes <rdar://problem/7696712> and part of PR 5255.
Anders Carlsson [Sat, 27 Feb 2010 20:02:53 +0000 (20:02 +0000)]
Enable the new vtable layout code for vtables that aren't construction vtables. (This doesn't mean that we emit LLVM IR using it yet, it just means that it's running and hopefully not crashing or asserting).
Douglas Gregor [Sat, 27 Feb 2010 02:42:25 +0000 (02:42 +0000)]
Robustify SourceManager::getLocation(), so that it returns an
end-of-line source location when given a column number beyond the
length of the line, or an end-of-file source location when given a
line number beyond the length of the file. Previously, we would return
an invalid location.
Douglas Gregor [Sat, 27 Feb 2010 01:32:48 +0000 (01:32 +0000)]
When given unsaved files in clang_createTranslationUnitFromSourceFile,
copy the source buffers provided rather than referencing them
directly, so that the caller can free those buffers immediately after
calling clang_createTranslationUnitFromSourceFile(). Otherwise, we
risk hitting those buffers later (when building source ranges, forming
diagnostics, etc.).
Ted Kremenek [Fri, 26 Feb 2010 19:18:41 +0000 (19:18 +0000)]
For printf format string checking, move the tracking of the data argument index out of
Sema and into analyze_printf::ParseFormatString(). Also use a bitvector to determine
what arguments have been covered (instead of just checking to see if the last argument consumed is the max argument). This is prep. for support positional arguments (an IEEE extension).
Zhongxing Xu [Fri, 26 Feb 2010 15:43:34 +0000 (15:43 +0000)]
Use a GDM to record the returned expression in the state when VisitReturnStmt.
Use this information to find the returned value and bind it to CallExpr in
ProcessCallExit.
And there is no need to remove dead bindings in ProcessCallExit, because
a. it would clean up the return value bound to CallExpr
b. we still would do it in the next ProcessStmt(), where we would not misclean
up the return value.
John McCall [Fri, 26 Feb 2010 08:45:28 +0000 (08:45 +0000)]
Fix an assertion-on-error during tentative constructor parsing by
propagating error conditions out of the various annotate-me-a-snowflake
routines. Generally (but not universally) removes redundant diagnostics
as well as, you know, not crashing on bad code. On the other hand,
I have just signed myself up to fix fiddly parser errors for the next
week. Again.
Douglas Gregor [Fri, 26 Feb 2010 06:03:23 +0000 (06:03 +0000)]
An explicit specialization is allowed following an explicit
instantiation so long as that explicit specialization was declared
previously. Fixes PR6160.
Douglas Gregor [Fri, 26 Feb 2010 05:06:18 +0000 (05:06 +0000)]
Implement semantic analysis for C++ [expr.new]p18-20, which describe
how we find the operator delete that matches withe operator new we
found in a C++ new-expression.
This will also need CodeGen support. On a happy note, we're now a
"nans" away from building tramp3d-v4.
Douglas Gregor [Fri, 26 Feb 2010 00:38:10 +0000 (00:38 +0000)]
Make sure to mark constructors, operator new, and operator delete as
used when we instantiate C++ new expressions, delete expressions, and
object-construction expressions. Fixes PR6424, although we can't test
all of it until we finish implementing lookup of "operator delete" for
new expressions (!).
Douglas Gregor [Fri, 26 Feb 2010 00:01:57 +0000 (00:01 +0000)]
When we decide to re-use an existing CXXConstructExpr node, make sure
to mark the constructor as referenced. Fixes the narrow issue reported
in PR6424, but there are a few other places that I'll fix before
closing out that PR.
Douglas Gregor [Thu, 25 Feb 2010 22:29:57 +0000 (22:29 +0000)]
When computing the composite pointer type for relational comparisons,
equality comparisons, and conditional operators, produce a composite
pointer type with the appropriate additional "const" qualifiers if the
pointer types would otherwise be incompatible. This is a small
extension (also present in GCC and EDG in a slightly different form)
that permits code like:
void** i; void const** j;
i == j;
with the following extwarn:
t.cpp:5:5: warning: comparison of distinct pointer types ('void **' and
'void const **') uses non-standard composite pointer type
'void const *const *' [-pedantic]
i == j;
~ ^ ~
Fixes PR6346, and I'll be filing a core issue about this with the C++
committee.
Douglas Gregor [Thu, 25 Feb 2010 19:01:05 +0000 (19:01 +0000)]
Allow us to compare derived-to-base conversions between a reference
binding and a copy-construction. Fixes an overloading problem in the
Clang-on-Clang build.
Gabor Greif [Thu, 25 Feb 2010 13:04:33 +0000 (13:04 +0000)]
Add "template" keyword at strategic position to fix
compilation using g++ v3.4.
I'll watch the buildbots and back out if necessary.
Feel free to do the same if something breaks.
Without this patch I get (on g++ 3.4.6) following error:
In file included from clang/lib/Sema/SemaTemplate.cpp:14:
clang/lib/Sema/TreeTransform.h: In member function `clang::ASTOwningResult<&clang::ActionBase::DeleteExpr> clang::TreeTransform<Derived>::RebuildCXXPseudoDestructorExpr(clang::ASTOwningResult<&clang::ActionBase::DeleteExpr>, clang::SourceLocation, bool, clang::NestedNameSpecifier*, clang::SourceRange, clang::TypeSourceInfo*, clang::SourceLocation, clang::SourceLocation, clang::PseudoDestructorTypeStorage)':
clang/lib/Sema/TreeTransform.h:5784: error: expected primary-expression before '>' token
clang/lib/Sema/TreeTransform.h:5784: error: expected primary-expression before ')' token
make[4]: *** [clang/lib/Sema/Release/SemaTemplate.o] Error 1
John McCall [Thu, 25 Feb 2010 10:46:05 +0000 (10:46 +0000)]
When comparing two method overload candidates during overload diagnostics,
skip the object argument conversion if either of the candidates didn't
initialize it.
Fixes PR6421, which is such a very straightforward extension of PR6398 that I
should have worked it into the last test case (and therefore caught it then).
Ah well.
Daniel Dunbar [Thu, 25 Feb 2010 08:49:05 +0000 (08:49 +0000)]
Add a minimal C interpreter example.
- Demonstrates how to build a standalone tool which loads source code using the
Driver and Frontend libraries, and then uses CodeGen and the JIT to actually
execute the code.
- Still more complicated than it should be, but hey its only 153 lines. :)
--
ddunbar@ozzy:tmp$ cat hello.c
#include <stdio.h>
int main() { printf("hello world\n"); return 0; }
ddunbar@ozzy:tmp$ clang-interpreter hello.c
hello world
--
Chandler Carruth [Thu, 25 Feb 2010 07:20:54 +0000 (07:20 +0000)]
Add a new conversion rank to classify conversions between complex and scalar
types. Rank these conversions below other conversions. This allows overload
resolution when the only distinction is between a complex and scalar type. It
also brings the complex overload resolutin in line with GCC's.
Ted Kremenek [Thu, 25 Feb 2010 05:44:09 +0000 (05:44 +0000)]
Add MacOSXAPIChecker, a meta checker to include various precondition checks for calls
to various MacOS X functions. The checks in BasicObjCFoundationChecks.cpp will
gradually be migrated here.
As a first check, check that when 'dispatch_once()' is passed a predicate value
that has non-local storage.
Douglas Gregor [Thu, 25 Feb 2010 04:46:04 +0000 (04:46 +0000)]
Restore the invariant that a nested-name-specifier can only contain
class types, dependent types, and namespaces. I had previously
weakened this invariant while working on parsing pseudo-destructor
expressions, but recent work in that area has made these changes
unnecessary.
Daniel Dunbar [Thu, 25 Feb 2010 04:37:45 +0000 (04:37 +0000)]
Frontend: Pull CodeGenAction out more, and eliminate CreateBackendConsumer.
This is the way I would like to move the frontend function towards -- distinct
pieces of functionality should be exposed only via FrontendAction
implementations which have clean and relatively-stable APIs.
This also isolates the surface area in clang which depends on LLVM CodeGen.
projects / clang / log