Pawel Wodnicki [Thu, 29 Nov 2012 14:56:07 +0000 (14:56 +0000)]
Merging r167918: into the 3.2 release branch.
When we have a MemberExpr referring to an overloaded static member function,
and we resolve it to a specific function based on the type which it's used as,
don't forget to mark it as referenced.
Pawel Wodnicki [Wed, 28 Nov 2012 23:44:46 +0000 (23:44 +0000)]
Merging r168818: into the 3.2 release branch.
PR13098: If we're instantiating an overloaded binary operator and we could
determine which member function would be the callee from within the template
definition, don't pass that function as a "non-member function" to
CreateOverloadedBinOp. Instead, just rely on it to select the member function
for itself.
Pawel Wodnicki [Mon, 26 Nov 2012 21:20:28 +0000 (21:20 +0000)]
Merging r167935: into the 3.2 release branch.
The ObjC++-to-C++ personality trick is only necessary on NeXT runtimes,
which is not coincidentally the only place it works, either (because
of how it tests for EH_TYPE symbols).
Pawel Wodnicki [Wed, 21 Nov 2012 08:33:40 +0000 (08:33 +0000)]
Merging r168063 into 3.2 release branch
block extended signatur option. Change previous option
to a cc1 -fencode-extended-block-signature and pass it
to cc1 and recognize this option to produce extended block
type signature. // rdar://12109031
Pawel Wodnicki [Wed, 21 Nov 2012 07:56:23 +0000 (07:56 +0000)]
Merging r168297 into 3.2 release branch
Completely re-work how the Clang driver interprets PIC and PIE options.
There were numerous issues here that were all entangled, and so I've
tried to do a general simplification of the logic.
1) The logic was mimicing actual GCC bugs, rather than "features". These
have been fixed in trunk GCC, and this fixes Clang as well. Notably,
the logic was always intended to be last-match-wins like any other
flag.
2) The logic for handling '-mdynamic-no-pic' was preposterously unclear.
It also allowed the use of this flag on non-Darwin platforms where it
has no actual meaning. Now this option is handled directly based on
tests of how llvm-gcc behaves, and it is only supported on Darwin.
3) The APIs for the Driver's ToolChains had the implementation ugliness
of dynamic-no-pic leaking through them. They also had the
implementation details of the LLVM relocation model flag names
leaking through.
4) The actual results of passing these flags was incorrect on Darwin in
many cases. For example, Darwin *always* uses PIC level 2 if it uses
in PIC level, and Darwin *always* uses PIC on 64-bit regardless of
the flags specified, including -fPIE. Darwin never compiles in PIE
mode, but it can *link* in PIE mode.
5) Also, PIC was not always being enabled even when PIE was. This isn't
a supported mode at all and may have caused some fallout in builds
with complex PIC and PIE interactions.
The result is (I hope) cleaner and clearer for readers. I've also left
comments and tests about some of the truly strage behavior that is
observed on Darwin platforms. We have no real testing of Windows
platforms and PIC, but I don't have the tools handy to figure that out.
Hopefully others can beef up our testing here.
Unfortunately, I can't test this for every platform. =/ If folks have
dependencies on these flags that aren't covered by tests, they may
break. I've audited and ensured that all the changes in behavior of the
existing tests are intentional and good. In particular I've tried to
make sure the Darwin behavior (which is more suprising than the Linux
behavior) also matches that of 'gcc' on my mac.
Pawel Wodnicki [Mon, 19 Nov 2012 21:01:40 +0000 (21:01 +0000)]
Merging r167884,r167920 from trunk into 3.2 release branch
r167884
Don't try to save the assigned value in a Objective-C property assignment
if the type of the value is a non-trivial class type. Fixes PR14318.
(There's a minor ObjC++ language change here: given that we can't save the
value, the type of the assignment expression is void in such cases.)
r167920
PR14279: Work around this major miscompilation by treating move operations as
non-trivial if they would not call a move operation, even if they would in fact
call a trivial copy operation. A proper fix is to follow, but this small
directed fix is intended for porting to the 3.2 release branch.
Anna Zaks [Tue, 13 Nov 2012 20:22:03 +0000 (20:22 +0000)]
Fix a Malloc Checker FP by tracking return values from initWithCharacter
and other functions.
When these functions return null, the pointer is not freed by
them/ownership is not transfered. So we should allow the user to free
the pointer by calling another function when the return value is NULL.
Anna Zaks [Tue, 13 Nov 2012 00:48:20 +0000 (00:48 +0000)]
[analyzer] Follow up to r167762 - precisely determine the adjustment
conditions.
The adjustment is needed only in case of dynamic dispatch performed by
the analyzer - when the runtime declaration is different from the static
one.
Document this explicitly in the code (by adding a helper). Also, use
canonical Decls to avoid matching against the case where the definition
is different from found declaration.
This fix suppresses the testcase I added in r167762, so add another
testcase to make sure we do test commit r167762.
Ted Kremenek [Mon, 12 Nov 2012 23:01:51 +0000 (23:01 +0000)]
Merge in r167749, per discussion on cfe-dev that we want -Wimplicit-fallthrough to only be active for C++11
until we come up with a language annotation for this warning that works outside of C++11.
Rafael Espindola [Mon, 12 Nov 2012 04:10:23 +0000 (04:10 +0000)]
In Sema::MergeVarDecl we handle merging of storage classes and visibility
attributes. In cases where the merged declaration is fully equivalent to the
two original ones, some of the code was getLVForDecl was duplicated.
Cases that are still handled in getLVForDecl are things like
__private_extern__ int N;
int N;
For which we cannot produce a single merged decl with all the information.
Daniel Jasper [Sun, 11 Nov 2012 22:14:55 +0000 (22:14 +0000)]
Fix binding of nodes in case of forEach..() matchers.
When recursively visiting the generated matches, the aggregated bindings need
to be copied during the recursion. Otherwise, we they might not be properly
overwritten (which is shown by the test), or there might be bound nodes present
that were bound on a different matching branch.
Douglas Gregor [Sat, 10 Nov 2012 07:24:09 +0000 (07:24 +0000)]
Rework my implementation of circular-reference finding to not use
CXXRecordDecl::forallBases, which does *not* do what I need. Fixes the
failure introduced in r167651.
Michael Liao [Sat, 10 Nov 2012 05:17:46 +0000 (05:17 +0000)]
Add clang support of RTM from TSX
- New options '-mrtm'/'-mno-rtm' are added to enable/disable RTM feature
- Builtin macro '__RTM__' is defined if RTM feature is enabled
- RTM intrinsic header is added and introduces 3 new intrinsics, namely
'_xbegin', '_xend', and '_xabort'.
- 3 new builtins are added to keep compatible with gcc, namely
'__builtin_ia32_xbegin', '__builtin_ia32_xend', and '__builtin_ia32_xabort'.
- Test cases for pre-defined macro and new intrinsic codegen are added.
Ted Kremenek [Sat, 10 Nov 2012 01:47:40 +0000 (01:47 +0000)]
Disable -Wimplicit-fallthrough when not using C++.
The rationale is that there is no good workflow to silence the warning
for specific cases, other than using pragmas. This is because the
attribute to decorate an explicit fall through is only available
in C++11.
By that argument, this should probably also be disabled unless one
is using C++11, but apparently there is an explicit test case for
this warning when using C++98. This will require further discussion
on cfe-commits.
Jordan Rose [Sat, 10 Nov 2012 01:40:08 +0000 (01:40 +0000)]
[analyzer] When invalidating symbolic offset regions, take fields into account.
Previously, RegionStore was being VERY conservative in saying that because
p[i].x and p[i].y have a concrete base region of 'p', they might overlap.
Now, we check the chain of fields back up to the base object and check if
they match.
This only kicks in when dealing with symbolic offset regions because
RegionStore's "base+offset" representation of concrete offset regions loses
all information about fields. In cases where all offsets are concrete
(s.x and s.y), RegionStore will already do the right thing, but mixing
concrete and symbolic offsets can cause bindings to be invalidated that
are known to not overlap (e.g. p[0].x and p[i].y).
This additional refinement is tracked by <rdar://problem/12676180>.
Douglas Gregor [Sat, 10 Nov 2012 01:18:17 +0000 (01:18 +0000)]
Diagnostic circular inheritance involving dependent base classes. We
would have diagnosed this at instantiation time anyway, if only we
didn't hang on all of these test cases. Fixes <rdar://problem/12629723>
David Blaikie [Sat, 10 Nov 2012 01:04:23 +0000 (01:04 +0000)]
PR14296: function parameter name collisions in function try/catch
C++11 3.3.3/2 "A parameter name shall not be redeclared in the outermost block
of the function definition nor in the outermost block of any handler associated
with a function-try-block."
It's not totally clear to me whether the "FIXME" case is covered by this, but
Richard Smith thinks it probably should be. It's just a bit more involved to
fix that case.
Richard Smith [Fri, 9 Nov 2012 23:03:14 +0000 (23:03 +0000)]
PR13788: Don't perform checks on the initializer of a dependently-typed
variable. Previously we didn't notice the type was dependent if the only
dependence came from an array bound.
Richard Smith [Fri, 9 Nov 2012 22:36:44 +0000 (22:36 +0000)]
PR14303: Add a NoDriverOption flag to those options which are not accepted by
the driver (the options defined in CC1Options.td) and exclude their help from
"clang --help".
Add a FileCharacteristic parameter to SourceManager::createFileIDForMemBuffer
for completeness and use it in CompilerInstance::InitializeSourceManager if
the input is a memory buffer.
Andy Gibbs [Fri, 9 Nov 2012 13:24:30 +0000 (13:24 +0000)]
Improved support for removing the comma preceding __VA_ARGS__ where __VA_ARGS__
is empty in a variadic macro expansion. This fixes a divergence in support for
the ", ## __VA_ARGS__" GCC extension which differed in behaviour when in strict
C99 mode (note: there is no change in behaviour has been made in the gnu99 mode
that clang uses by default). In addition, there is improved support for the
Microsoft alternative extension ", __VA_ARGS__".
Fix a bug I found while preparing my devmtg talk: When passing NULL to a
function that takes a const Foo&, where Foo is convertible from a large number
of pointer types, we print ALL the overloads, no matter the setting of
-fshow-overloads.
There is potential follow-on work in unifying the "print candidates, but not
too many" logic between OverloadCandidateSet::NoteCandidates and
ImplicitConversionSequence::DiagnoseAmbiguousConversion.
Benjamin Kramer [Thu, 8 Nov 2012 12:59:15 +0000 (12:59 +0000)]
Set TLSSupported to false on Haiku.
Haiku does not support this (yet). Leaving it set to true leads to
configure scripts detecting __thread being available and Clang emitting
code for it, resulting in binaries the runtime_loader will refuse to
load.
Richard Smith [Wed, 7 Nov 2012 23:56:21 +0000 (23:56 +0000)]
When deciding whether to convert an array construction loop into a memcpy, look
at whether the *selected* constructor would be trivial rather than considering
whether the array's element type has *any* non-trivial constructors of the
relevant kind.
Dmitri Gribenko [Wed, 7 Nov 2012 23:52:25 +0000 (23:52 +0000)]
Force C89 for c-index-test.c
MSVC supports only C89, so it is important to keep c-index-test.c
buildable with C89. However, Clang defaults to C99, so while building
Clang with Clang one can introduce C99 constructs into c-index-test.c
without noticing.
Thanks to Nakamura Takumi for helping with MSVC bits.
Richard Trieu [Wed, 7 Nov 2012 21:17:13 +0000 (21:17 +0000)]
When template deduction fails on a derived class, try a template deduction on
the base class. If the base class deduction succeeds, use those results. If
it fails, keep using the results from the derived class template deduction.
This prevents an assertion later where the type of deduction failure doesn't
match up with the template deduction info.
projects / clang / log