Douglas Gregor [Sat, 24 Apr 2010 15:35:55 +0000 (15:35 +0000)]
Keep track of when DependentNameTypes have no associated keyword
(e.g., no typename, enum, class, etc.), e.g., because the context is
one that is known to refer to a type. Patch from Enea Zaffanella!
Dan Gohman [Sat, 24 Apr 2010 04:55:02 +0000 (04:55 +0000)]
Fix a place in inline asm lowering which was creating a TruncInst with a
pointer operand. This fixes an abort on
MultiSource/Applications/ClamAV/libclamav_mbox.c.
John McCall [Sat, 24 Apr 2010 01:30:58 +0000 (01:30 +0000)]
Recommit r102215, this time being more careful to only set the "principal
declaration" (i.e. the only which will actually be looked up) to have the
non-member-operator bit.
Douglas Gregor [Fri, 23 Apr 2010 23:01:43 +0000 (23:01 +0000)]
Rework Parser-Sema interface for Objective-C @catch exception object
arguments. Rather than having the parser call ActOnParamDeclarator
(which is a bit of a hack), call a new ActOnObjCExceptionDecl
action. We'll be moving more functionality into this handler to
perform earlier checking of @catch.
Douglas Gregor [Fri, 23 Apr 2010 22:50:49 +0000 (22:50 +0000)]
Improve the AST representation of Objective-C @try/@catch/@finally
statements. Instead of the @try having a single @catch, where all of
the @catch's were chained (using an O(n^2) algorithm nonetheless),
@try just holds an array of its @catch blocks. The resulting AST is
slightly more compact (not important) and better represents the actual
language semantics (good).
Anders Carlsson [Fri, 23 Apr 2010 20:40:38 +0000 (20:40 +0000)]
Add an inheritance path member variable to CastExpr. For now it's always null but for derived-to-base and base-to-derived cast expressions it will contain the full base path. This is needed to avoid ambiguities.
Daniel Dunbar [Fri, 23 Apr 2010 19:12:32 +0000 (19:12 +0000)]
NeXT/EH: When generating the rethrow code for a finally block, make sure to
chain outwards when inside a nested exception scope.
- A real test for this is going into LLVM test-suite.
Chris Lattner [Fri, 23 Apr 2010 17:27:29 +0000 (17:27 +0000)]
Implement PR6845. We allow matching constraints to have different
input and output types when the smaller value isn't mentioned in the
asm string. Extend this support from integers to also allowing
fp values to be mismatched (if not mentioned in the asm string).
Douglas Gregor [Fri, 23 Apr 2010 16:25:07 +0000 (16:25 +0000)]
When instantiating a typedef of an anonymous tag type, note in the tag
declaration that this typedef gives the tag a name. Fixes a problem
uncovered by Boost.GIL (Generic Image Library).
Douglas Gregor [Fri, 23 Apr 2010 04:16:32 +0000 (04:16 +0000)]
Handle compound assignment expressions (i += j) as lvalues, which is
permitted in C++ but not in C. Fixes PR6900. Clang can now handle all
of Boost.Lambda's regression tests.
John McCall [Fri, 23 Apr 2010 02:41:41 +0000 (02:41 +0000)]
C++ doesn't really use "namespaces" for different kinds of names the same
way that C does. Among other differences, elaborated type specifiers
are defined to skip "non-types", which, as you might imagine, does not
include typedefs. Rework our use of IDNS masks to capture the semantics
of different kinds of declarations better, and remove most current lookup
filters. Removing the last remaining filter is more complicated and will
happen in a separate patch.
Fixes PR 6885 as well some spectrum of unfiled bugs.
Douglas Gregor [Fri, 23 Apr 2010 02:08:13 +0000 (02:08 +0000)]
When parsing a cast-expression that starts with a scope annotation,
try to annotate as a type first to determine whether we have a
functional-style cast. Patch by Eli Friedman, fixes PR6830.
Douglas Gregor [Thu, 22 Apr 2010 23:19:50 +0000 (23:19 +0000)]
Kill off IDNS_ObjCImplementation and IDNS_ObjCCategoryName; they
aren't and never were used. There's a gap in the bit pattern for IDNS
now, but I'm sure *someone* will fill it.
John McCall [Thu, 22 Apr 2010 18:44:12 +0000 (18:44 +0000)]
Use the naming class from the overloaded lookup when access-checking an
address of overloaded function, instead of assuming that a nested name
specifier was used. A nested name specifier is not required for static
functions.
Douglas Gregor [Thu, 22 Apr 2010 17:28:13 +0000 (17:28 +0000)]
Some Objective-C++ types and expressions will never change during
template instantiation, since they cannot be dependent or have
dependent parts. Handle them the simple way.
Douglas Gregor [Thu, 22 Apr 2010 17:01:48 +0000 (17:01 +0000)]
When a dependent Objective-C++ message send was able to resolve the
method being called at template definition time, retain that method
and pass it through to type-checking. We will not perform any lookup
for the method during template instantiation.
Douglas Gregor [Thu, 22 Apr 2010 16:50:51 +0000 (16:50 +0000)]
Remove the SelectorLoc argument to Sema::BuildInstanceMesssage and
Sema::BuildClassMessage; we weren't using it, and template
instantiation was faking it anyway.
Douglas Gregor [Thu, 22 Apr 2010 16:44:27 +0000 (16:44 +0000)]
Implement template instantiation for Objective-C++ message sends. We
support dependent receivers for class and instance messages, along
with dependent message arguments (of course), and check as much as we
can at template definition time.
This commit also deals with a subtle aspect of template instantiation
in Objective-C++, where the type 'T *' can morph from a dependent
PointerType into a non-dependent ObjCObjectPointer type.
Daniel Dunbar [Thu, 22 Apr 2010 16:14:54 +0000 (16:14 +0000)]
ARM/APCS: Don't respect bit-field types when laying out structures.
- This fixes the last known ABI issues with ARM/APCS.
- I've run the first 1k ABITests with '--no-unsigned --no-vector --no-complex'
on {armv6, armv7} x {-mno-thumb, -mthumb}, and the first 10k tests for armv7
-mthumb, for both function return types and single argument calls. These all
pass now (they failed horribly before without --no-bitfield).
Daniel Dunbar [Thu, 22 Apr 2010 15:22:33 +0000 (15:22 +0000)]
IRgen: Fix another case where we generated an invalid access component when we
immediately narrowed the access size. Fix this (and previous case) by just
choosing a better access size up-front.
Daniel Dunbar [Thu, 22 Apr 2010 14:56:10 +0000 (14:56 +0000)]
IRgen: Fix case where we might generate an access component with width == 0, if
we have to narrow the access side immediately (can happen with packed,
-fno-bitfield-type-align).
Douglas Gregor [Thu, 22 Apr 2010 14:36:26 +0000 (14:36 +0000)]
When checking whether to diagnose an initialized "extern" variable,
look for the const on the base type rather than on the top-level
type. Fixes PR6495 properly.
Daniel Dunbar [Thu, 22 Apr 2010 02:35:46 +0000 (02:35 +0000)]
IRgen: Rewrite bit-field access policy to not access data beyond the bounds of the structure, which we also now verify as part of the post-layout consistency checks.
- This fixes some pedantic bugs with packed structures, as well as major problems with -fno-bitfield-type-align.
- Fixes PR5591, PR5567, and all known -fno-bitfield-type-align issues.
John McCall [Thu, 22 Apr 2010 01:10:34 +0000 (01:10 +0000)]
Call PerformCopyInitialization to properly initialize the exception temporary
in a throw expression. Use EmitAnyExprToMem to emit the throw expression,
which magically elides the final copy-constructor call (which raises a new
strict-compliance bug, but baby steps). Give __cxa_throw a destructor pointer
if the exception type has a non-trivial destructor.
Douglas Gregor [Thu, 22 Apr 2010 00:20:18 +0000 (00:20 +0000)]
Whenever we complain about a failed initialization of a function or
method parameter, provide a note pointing at the parameter itself so
the user does not have to manually look for the function/method being
called and match up parameters to arguments. For example, we now get:
t.c:4:5: warning: incompatible pointer types passing 'long *' to
parameter of
type 'int *' [-pedantic]
f(long_ptr);
^~~~~~~~
t.c:1:13: note: passing argument to parameter 'x' here
void f(int *x);
^
Dan Gohman [Wed, 21 Apr 2010 23:32:43 +0000 (23:32 +0000)]
When computing the alignof value for a vector type, ensure that
the alignment is a power of 2, even in the esoteric case of a
vector element which does not have a power-of-2 sizeof value.
Douglas Gregor [Wed, 21 Apr 2010 23:24:10 +0000 (23:24 +0000)]
Switch the initialization of Objective-C message parameters (as occurs
during message sends) over to the new initialization code and away
from the C-only CheckSingleAssignmentConstraints. The enables the use
of C++ types in method parameters and message arguments, as well as
unifying more initialiation code overall.
Douglas Gregor [Wed, 21 Apr 2010 22:36:40 +0000 (22:36 +0000)]
Implement parsing for message sends in Objective-C++. Message sends in
Objective-C++ have a more complex grammar than in Objective-C
(surprise!), because
(1) The receiver of an instance message can be a qualified name such
as ::I or identity<I>::type.
(2) Expressions in C++ can start with a type.
The receiver grammar isn't actually ambiguous; it just takes a bit of
work to parse past the type before deciding whether we have a type or
expression. We do this in two places within the grammar: once for
message sends and once when we're determining whether a []'d clause in
an initializer list is a message send or a C99 designated initializer.
This implementation of Objective-C++ message sends contains one known
extension beyond GCC's implementation, which is to permit a
typename-specifier as the receiver type for a class message, e.g.,
[typename compute_receiver_type<T>::type method];
Note that the same effect can be achieved in GCC by way of a typedef,
e.g.,
Douglas Gregor [Wed, 21 Apr 2010 20:38:13 +0000 (20:38 +0000)]
Migrate the responsibility for turning the receiver name in an
Objective-C class message expression into a type from the parser
(which was doing so in two places) to Action::getObjCMessageKind()
which, in the case of Sema, reduces the number of name lookups we need
to perform.
Douglas Gregor [Wed, 21 Apr 2010 19:57:20 +0000 (19:57 +0000)]
Rework the Parser-Sema interaction for Objective-C message
sends. Major changes include:
- Expanded the interface from two actions (ActOnInstanceMessage,
ActOnClassMessage), where ActOnClassMessage also handled sends to
"super" by checking whether the identifier was "super", to three
actions (ActOnInstanceMessage, ActOnClassMessage,
ActOnSuperMessage). Code completion has the same changes.
- The parser now resolves the type to which we are sending a class
message, so ActOnClassMessage now accepts a TypeTy* (rather than
an IdentifierInfo *). This opens the door to more interesting
types (for Objective-C++ support).
- Split ActOnInstanceMessage and ActOnClassMessage into parser
action functions (with their original names) and semantic
functions (BuildInstanceMessage and BuildClassMessage,
respectively). At present, this split is onyl used by
ActOnSuperMessage, which decides which kind of super message it
has and forwards to the appropriate Build*Message. In the future,
Build*Message will be used by template instantiation.
- Use getObjCMessageKind() within the disambiguation of Objective-C
message sends vs. array designators.
Two notes about substandard bits in this patch:
- There is some redundancy in the code in ParseObjCMessageExpr and
ParseInitializerWithPotentialDesignator; this will be addressed
shortly by centralizing the mapping from identifiers to type names
for the message receiver.
- There is some #if 0'd code that won't likely ever be used---it
handles the use of 'super' in methods whose class does not have a
superclass---but could be used to model GCC's behavior more
closely. This code will die in my next check-in, but I want it in
Subversion.
John McCall [Wed, 21 Apr 2010 11:18:06 +0000 (11:18 +0000)]
I failed to notice that my last patch wasn't doing as much as it could
because EmitBranch actually clears the insert point. This version
actually accomplishes what I initially wanted.
John McCall [Wed, 21 Apr 2010 10:29:06 +0000 (10:29 +0000)]
Teach EmitBlock to put the target block immediately after the current block
(if there's a current block). The chief advantage of doing this is that it
lets us pick blocks (e.g. EH blocks) to push to the end of the function so
that fallthrough happens consistently --- i.e. it gives us the flexibility
of ordering blocks as we please without having to change the order in which
we generate code. There are standard (?) optimization passes which can do some
of that for us, but better to generate reasonable code to begin with.
projects / clang / log