Daniel Dunbar [Tue, 9 Sep 2008 01:06:48 +0000 (01:06 +0000)]
Change CodeGen to emit calls using (RValue,Type) list:
- Add CodeGenFunction::EmitAnyExprToTemp
o Like EmitAnyExpr, but emits aggregates to a temporary location if
none is available. Seems like this should be simpler (even aside
from using first class aggregates).
- Killed CodeGenFunction::EmitCallArg (just append the pair)
- Conversion of RValues to actual call arguments is now isolated in
CodeGenFunction::EmitCall.
Daniel Dunbar [Mon, 8 Sep 2008 23:44:31 +0000 (23:44 +0000)]
Fix a number of issues w.r.t. emission of global for functions and
aliases.
- Attributes specific to a definition are only set when the
definition is seen.
- Alias generation is delayed until the end of the module; necessary
since the alias may reference forward.
- Fixes: PR2743, <rdr://6140807&6094512>
- Improves: <rdr://6095112> (added XFAIL)
Daniel Dunbar [Mon, 8 Sep 2008 21:33:45 +0000 (21:33 +0000)]
Refactor parameter attribute handling:
- Add CGCall.h for dealing with ABI issues related to calls.
- Add CGFunctionInfo and CGCallInfo for capturing ABI relevant
information about functions and calls.
- Isolate LLVM parameter attribute handling inside CGCall.cpp
skip test if llvm-gcc is requires but not found on the path.
someone with llvm-gcc installed please test if the Codegen/function-attributes.c test isn't skip in your system. thanks.
Eli Friedman [Sat, 6 Sep 2008 01:37:51 +0000 (01:37 +0000)]
Per PR2773, define __USER_LABEL_PREFIX__ for x86-32 Linux and Windows.
If you're on some other platform, the correct definition for this macro
would be appreciated; to find the correct definition, just run the
following command:
Change struct forward declarations and definitions to use unique RecordDecls, as opposed to creating a single RecordDecl and reusing it.
This change effects both RecordDecls and CXXRecordDecls, but does not effect EnumDecls (yet).
The motivation of this patch is as follows:
- Capture more source information, necessary for refactoring/rewriting clients.
- Pave the way to resolve ownership issues with RecordDecls with the forthcoming
addition of DeclGroups.
Current caveats:
- Until DeclGroups are in place, we will leak RecordDecls not explicitly
referenced by the AST. For example:
typedef struct { ... } x;
The RecordDecl for the struct will be leaked because the TypedefDecl doesn't
refer to it. This will be solved with DeclGroups.
- This patch also (temporarily) breaks CodeGen. More below.
High-level changes:
- As before, TagType still refers to a TagDecl, but it doesn't own it. When
a struct/union/class is first referenced, a RecordType and RecordDecl are
created for it, and the RecordType refers to that RecordDecl. Later, if
a new RecordDecl is created, the pointer to a RecordDecl in RecordType is
updated to point to the RecordDecl that defines the struct/union/class.
- TagDecl and RecordDecl now how a method 'getDefinition()' to return the
TagDecl*/RecordDecl* that refers to the TagDecl* that defines a particular
enum/struct/class/union. This is useful from going from a RecordDecl* that
defines a forward declaration to the RecordDecl* that provides the actual
definition. Note that this also works for EnumDecls, except that in this case
there is no distinction between forward declarations and definitions (yet).
- Clients should no longer assume that 'isDefinition()' returns true from a
RecordDecl if the corresponding struct/union/class has been defined.
isDefinition() only returns true if a particular RecordDecl is the defining
Decl. Use 'getDefinition()' instead to determine if a struct has been defined.
- The main changes to Sema happen in ActOnTag. To make the changes more
incremental, I split off the processing of enums and structs et al into two
code paths. Enums use the original code path (which is in ActOnTag) and
structs use the ActOnTagStruct. Eventually the two code paths will be merged,
but the idea was to preserve the original logic both for comparison and not to
change the logic for both enums and structs all at once.
- There is NO CHAINING of RecordDecls for the same RecordType. All RecordDecls
that correspond to the same type simply have a pointer to that type. If we
need to figure out what are all the RecordDecls for a given type we can build
a backmap.
- The diff in CXXRecordDecl.[cpp,h] is actually very small; it just mimics the
changes to RecordDecl. For some reason 'svn' marks the entire file as changed.
Why is CodeGen broken:
- Codegen assumes that there is an equivalence between RecordDecl* and
RecordType*. This was true before because we only created one RecordDecl* for
a given RecordType*, but it is no longer true. I believe this shouldn't be too
hard to change, but the patch was big enough as it is.
I have tested this patch on both the clang test suite, and by running the static analyzer over Postgresql and a large Apple-internal project (mix of Objective-C and C).
Daniel Dunbar [Fri, 5 Sep 2008 03:22:57 +0000 (03:22 +0000)]
Workaround gcc bug causing crash on our preprocessed outputs.
- gcc is not happy if we start a preprocessed file with
#line 1 "XXX" 1
- Workaround by making sure file starts with a simple #line change.
Also, factored WriteLineInfo out.
Also, fixed bug where FileType was not being correctly updated.
Remove "NextDecl" from RecordDecl. This change touches many files that where RecordDecl or CXXRecordDecl was constructed, always with an argument of 'NULL' for the previous declaration.
The motivation behind this change is that chaining the RecordDecls is simply unnecessary. Once we create multiple RecordDecls for the same struct/union/class, clients that care about all the declarations of the same struct can build a back map by seeing which Decls refer to the same RecordType.
fix running tests with valgrind (there were a lot of bogus failures and warnings)
currently clang passes all tests under valgrind with the leak checker disabled :P (and fails most otherwise)
scan-build:
- Only set the environment variable 'CXX' if the user specifies --use-c++.
- Fix regression when setting LDPLUSPLUS: add a 'which' to determine the location of g++. This regression was pointed out by Jordan Breeding!
Steve Naroff [Thu, 4 Sep 2008 16:56:14 +0000 (16:56 +0000)]
Touchup CheckSingleAssignmentConstraints() and CheckCompareOperands() to check for block pointers.
Added a couple FIXME's wrt PointLikeType. If the author reads this, it would be great to get some background on this class (thanks in advance).
Daniel Dunbar [Thu, 4 Sep 2008 04:36:23 +0000 (04:36 +0000)]
Add some Objective-C code generation tests.
- Note that these don't really test anything other than that code
generation doesn't fail or crash. Better than nothing though!
Daniel Dunbar [Wed, 3 Sep 2008 17:53:25 +0000 (17:53 +0000)]
Improve type-checking of ?: for Objective-C types.
- Allow any Objective-C object types to devolve to type id in a ?:
expression. This matches gcc behavior more closely.
Store: (static analyzer)
- Change definition of store::Region and store::Binding (once again) to make
them real classes that just wrap pointers. This makes them more strictly
typed, and allows specific implementations of Regions/Bindings to just
subclass them.
- minor renamings to RegionExtent and its subclasses
- added a bunch of doxygen comments
StoreManager: (static analyzer)
- added 'iterBindings', an iteration method for iterating over the bindings of a
store. It that takes a callback object (acting like a poor man's closure).
- added 'getRVal' version for store::Binding. Will potentially phase the other
versions of GetRVal in StoreManager out.
- reimplemented 'getBindings' to be non-virtual and to use 'iterBindings'
BasicStoreManager: (static analyzer)
- implemented 'iterBindings' for BasicStoreManager
Daniel Dunbar [Wed, 3 Sep 2008 01:05:41 +0000 (01:05 +0000)]
Restore Objective-C dot-syntax access of methods.
- Now also searches for correct setter method.
- There are still some issues regarding validation of the setter
method and access of read-only properties.
RecordDecl:
- Remove method 'isForwardDecl'; this functionality is already provided by
'isDefinition()'
- Move method definitions to be co-located with other RecordDecl methods.
RecordDecl:
- Added method 'isForwardDeclaration', a predicate method that returns true
if a RecordDecl represents a forward declaration.
- Added method 'getDefinitionDecl', a query method that returns a pointer to
the RecordDecl that provides the actual definition of a struct/union.
CXXRecordDecl and RecordDecl:
- Change constructor and create methods to accept a CXXRecordDecl* (RecordDecl*)
instead of a ScopedDecl* for PrevDecl. This causes the type checking
to be more tight and doesn't break any code.
RecordDecl:
- Don't use the NextDeclarator field in ScopedDecl to represent the previous
declaration. This is a conflated use of the NextDeclarator field, which will
be removed anyway when DeclGroups are fully implemented.
- Instead, represent (a soon to be implemented) chain of RecordDecls using a
NextDecl field. The last RecordDecl in the chain is always the 'defining'
RecordDecl that owns the FieldDecls. The other RecordDecls in the chain
are forward declarations.
Steve Naroff [Tue, 2 Sep 2008 18:50:17 +0000 (18:50 +0000)]
- Implement __block.
- Replace FIXME in Preprocessor::HandleIdentifier() with a check that avoids diagnosing extension tokens that originate from macro definitions.
Eli Friedman [Tue, 2 Sep 2008 05:29:22 +0000 (05:29 +0000)]
Fix for PR2750; don't check for an 'e' in the trash after the token.
Note that this isn't really a complete fix; I think there are other
potential overrun situations. I don't really know what the best
systematic fix is, though.
Eli Friedman [Tue, 2 Sep 2008 05:19:23 +0000 (05:19 +0000)]
An extremely hacky version of transparent_union support; it isn't
anywhere near correct in terms of missing cases and missing
diagnostics, but it's good enough to handle the uses in the
Linux system headers, which are currently a constant pain for compiling
applications on Linux.
Eli Friedman [Tue, 2 Sep 2008 05:09:35 +0000 (05:09 +0000)]
Make sure to take the unqualified versions of the canonical types for
type-checking pointer subtraction; if the canonical types aren't used,
the qualifiers won't always get stripped off correctly.
Tidy up sema processing of attribute "nonull":
- warn about nonnull being applied to functions with no pointer arguments
- continue processing argument list in the attribute when we encounter a non-pointer parameter being marked as nonnull
- when no argument list is specified, only mark pointers as nonnull. This fixes PR 2732 and radar 6188814.
Daniel Dunbar [Sat, 30 Aug 2008 05:35:15 +0000 (05:35 +0000)]
Add Objective-C property setter support.
- Change Obj-C runtime message API, drop the ObjCMessageExpr arg in
favor of just result type and selector. Necessary so it can be
reused in situations where we don't want to cons up an
ObjCMessageExpr.
- Update aggregate binary assignment to know about special property
ref lvalues.
- Add CodeGenFunction::EmitCallArg overload which takes an already
emitted rvalue.
Add CodeGenFunction::StoreComplexIntoAddr.
Disabled logic in Sema for parsing Objective-C dot-syntax that
accesses methods. This code does not search in the correct order and
the AST node has no way of properly representing its results.
Updated StmtDumper to print a bit more information about
ObjCPropertyRefExprs.
Daniel Dunbar [Sat, 30 Aug 2008 03:02:31 +0000 (03:02 +0000)]
Refactor handling of calls:
- Added CodeGenFunction::EmitCall which just takes the callee, return
type, and a list of (Value*,QualType) pairs.
- Added CodeGenFunction::EmitCallArg which handles emitting code for
a call argument and turning it into an appropriate
(Value*,QualType) pair.
- Changed Objective-C runtime interface so that the actual emission
of arguments for message sends is (once again) done in the code to
emit a message send.
No intended functionality change, this is prep work for better ABI
support and for Objective-C property setter support.
Daniel Dunbar [Fri, 29 Aug 2008 21:03:27 +0000 (21:03 +0000)]
Improve ccc:
- Support environment variable CCC_LANGUAGES to control which
languages clang is invoked on. If unset clang is invoked for all
languages, otherwise CCC_LANGUAGES should be a comma separated list
of the languages (as accepted by -x) for which clang should be
invoked. Useful for only building C and Objective-C parts of a
project with clang, for example.
- Add environment variable CCC_FALLBACK. If set and non-empty then
ccc will try and compile using the regular compiler if compilation
with clang fails.
- A few other tweaks to add options, flush stdout, recognize .mm as
objective-c++, and infer languages for compile+link style
invocations.
projects / clang / log