Nick Lewycky [Fri, 11 Oct 2013 03:35:10 +0000 (03:35 +0000)]
Add -fno-function-sections and -fno-data-sections. Since
-f{function,data}-sections had no tests at all, add some, and verify that the
-fno variants work as well.
ObjectiveC. ObjectiveC's collection selector expression in
the fereach loop must be a non-const lvalue expression as
it will be assigned to at the beginning of the loop.
// rdar://15123684
Hans Wennborg [Thu, 10 Oct 2013 18:03:08 +0000 (18:03 +0000)]
Locate VS InstallDir in the presence of newer runtime
This fixes getSystemRegistryString() in WindowsToolChain.cpp to
make sure that the VS version that it picks has an InstallDir.
Previously we would look for the highest version os VS and check
for InstallDir afterwards.
Hao Liu [Thu, 10 Oct 2013 14:59:36 +0000 (14:59 +0000)]
Implement AArch64 vector load/store multiple N-element structure class SIMD(lselem).
Including following 14 instructions:
4 ld1 insts: load multiple 1-element structure to sequential 1/2/3/4 registers.
ld2/ld3/ld4: load multiple N-element structure to sequential N registers (N=2,3,4).
4 st1 insts: store multiple 1-element structure from sequential 1/2/3/4 registers.
st2/st3/st4: store multiple N-element structure from sequential N registers (N = 2,3,4).
E.g. ld1(3 registers version) will load 32-bit elements {A, B, C, D, E, F} sequentially into the three 64-bit vectors list {BA, DC, FE}.
E.g. ld3 will load 32-bit elements {A, B, C, D, E, F} into the three 64-bit vectors list {DA, EB, FC}.
Ted Kremenek [Thu, 10 Oct 2013 00:54:01 +0000 (00:54 +0000)]
Fix getIntegerTypeOrder() to properly handle enums by first unwrapping their underlying integer type. This is a precondition for calling getIntegerRank().
Fixes an assertion failure in a test case involving vectors.
Hans Wennborg [Wed, 9 Oct 2013 23:41:48 +0000 (23:41 +0000)]
Use KEY_WOW64_32KEY when reading the registry in WindowsToolChain.cpp (PR17033)
This exposes a 32-bit view of the registry even when Clang is built as a 64-bit
program. Since Visual Studio is a 32-bit application, this is necessary for us
to find it.
Ted Kremenek [Wed, 9 Oct 2013 22:34:33 +0000 (22:34 +0000)]
Refine string literal concatenation warning within an NSArray literal to not warn when the literal comes from a macro expansion. Fixes <rdar://problem/15147688>.
Sean Callanan [Wed, 9 Oct 2013 21:45:11 +0000 (21:45 +0000)]
This patch addresses a problem encountered by the
ASTImporter when importing the following types:
typedef struct {
} A;
typedef struct {
A a;
} B;
Suppose we have imported B, but we did not at that
time need to complete it. Then later we want to
import A. The struct is anonymous, so the first
thing we want to do is make sure no other anonymous
struct already matches it. So we set up an
StructuralEquivalenceContext and compare B with A.
This happens at ASTImporter.cpp:2179.
Now, in this scenario, B is not complete. So we go
and import its fields, including a, which causes A
to be imported. The ASTImporter doesn’t yet have A
in its list of already-imported things, so we
import A.
After the StructuralEquivalenceContext is finished
determining that A and B are different, the
ASTImporter concludes that A must be imported
because no equivalent exists, so it imports a second
copy of A. Now we have two different structs
representing A. This is really bad news.
The patch allows the StructuralEquivalenceContext to
use the original version of B when making its
comparison, obviating the need for an import and
cutting this loop.
Consumed analysis: improve loop handling. The prior version of the analysis
marked all variables as "unknown" at the start of a loop. The new version
keeps the initial state of variables unchanged, but issues a warning if the
state at the end of the loop is different from the state at the beginning.
This patch will eventually be replaced with a more precise analysis.
Initial patch by chris.wailes@gmail.com. Reviewed and edited by
delesley@google.com.
Hans Wennborg [Wed, 9 Oct 2013 18:10:25 +0000 (18:10 +0000)]
Tighten diagnostics for calling conventions on variadic functions
Follow-up from r192240.
This makes it an error to use callee-cleanup conventions on variadic
functions, except for __fastcall and __stdcall, which we ignore with
a warning for GCC and MSVC compatibility.
Reid Kleckner [Tue, 8 Oct 2013 22:45:29 +0000 (22:45 +0000)]
-Wmicrosoft: Don't warn on non-inline pure virtual method definitions
MSVC and clang with -fms-extensions allow pure virtual methods to be
defined inline after the "= 0" tokens. Clang warns on these because it
is not standard, but incorrectly warns on out-of-line definitions, which
are standard.
With this change, clang will only warn on inline definitions of pure
virtual methods.
Fixes some self-host warnings on out-of-line definitions of pure virtual
destructors.
Hans Wennborg [Tue, 8 Oct 2013 21:52:56 +0000 (21:52 +0000)]
Turn error about fastcall variadic function into warning in MS mode (PR12535)
MSVC allows this and silently falls back to __cdecl for variadic functions.
This patch turns Clang's error into a warning in MS mode and adds a test
to make sure we generate correct code.
ObjectiveC migration. Add support for inferring
properties of block pointer types. Also, remove
strong lifetime attribute from property type
in this migration. This is wip.
// rdar://15082818
Benjamin Kramer [Tue, 8 Oct 2013 16:58:52 +0000 (16:58 +0000)]
Fix an edge case in the template differ with default arguments.
In the test case one type is coming from a typedef with no default arg, the
other has the default arg. Taking the default arg from the typedef crashes, so
always use the real template paramter declaration. PR17510.
Daniel Jasper [Tue, 8 Oct 2013 15:54:36 +0000 (15:54 +0000)]
clang-format: Don't exit with failure on empty files.
Also let clang-format-diff.py detect errors based on clang-format's
return code. Otherwise messages like "Can't find usable .clang-format,
falling back to LLVM style" can make it fail, which might be undesired.
Reid Kleckner [Tue, 8 Oct 2013 00:58:57 +0000 (00:58 +0000)]
[ms-cxxabi] Fix the calling convention for operator new in records
Summary:
Operator new, new[], delete, and delete[] are all implicitly static when
declared inside a record. CXXMethodDecl already knows this, but we need
to account for that before we pick the calling convention for the
function type.
Justin Bogner [Tue, 8 Oct 2013 00:19:09 +0000 (00:19 +0000)]
Sema: Only merge typedef attributes if the previous decl is a typedef
In r186373, we started merging attributes on typedefs, but this causes
us to try to merge attributes even if the previous declaration was not
a typedef.
Only merge the attributes if the previous decl was also a typedef.
Manman Ren [Tue, 8 Oct 2013 00:08:49 +0000 (00:08 +0000)]
TBAA: use the same format for scalar TBAA and struct-path aware TBAA.
An updated version of r191586 with bug fix.
Struct-path aware TBAA generates tags to specify the access path,
while scalar TBAA only generates tags to scalar types.
We should not generate a TBAA tag with null being the first field. When
a TBAA type node is null, the tag should be null too. Make sure we
don't decorate an instruction with a null TBAA tag.
Added a testing case for the bug reported by Richard with -relaxed-aliasing
and -fsanitizer=thread.
I'm not completely satisfied with this patch. Sprinkling "diagnostic ignored"
_Pragmas throughout this file is gross, but I couldn't suppress
it for the entire file.
ObjectiveC migrator: A typical implementation of
'default' methods in Foundation does not
infer 'instancetype' for methods' result type.
// rdar://15145218
Richard Smith [Mon, 7 Oct 2013 19:57:58 +0000 (19:57 +0000)]
Add support for WG21 N3599 (literal operator template for strings) as a GNU
extension. The GCC folks have decided to support this even though the standard
committee have not yet approved this feature.
Richard Smith [Mon, 7 Oct 2013 08:02:11 +0000 (08:02 +0000)]
When merging class definitions across modules in C++, merge together fields.
This change doesn't go all the way to making fields redeclarable; instead, it
makes them 'mergeable', which means we can find the canonical declaration, but
not much else (and for a declaration that's not from a module, the canonical
declaration is always that declaration).
David Majnemer [Mon, 7 Oct 2013 07:33:27 +0000 (07:33 +0000)]
Driver: Use the canonical command line arguments.
Summary:
Use the arguments given to the OS at process creation-time instead of
the arguments passed into main() by the C runtime environment. The ones
that main() received may not be suitable (e.g. not Unicode).
Mark Lacey [Sun, 6 Oct 2013 01:33:34 +0000 (01:33 +0000)]
Pass CGCXXABIs around directly.
In functions that only need to use the CGCXXABI member of a CodeGenTypes
class, pass that reference around directly rather than a reference to
a CodeGenTypes class.
This makes the actual dependence on CGCXXABI clear at the call sites.