Douglas Gregor [Wed, 21 Apr 2010 00:45:42 +0000 (00:45 +0000)]
Overhaul the AST representation of Objective-C message send
expressions, to improve source-location information, clarify the
actual receiver of the message, and pave the way for proper C++
support. The ObjCMessageExpr node represents four different kinds of
message sends in a single AST node:
1) Send to a object instance described by an expression (e.g., [x method:5])
2) Send to a class described by the class name (e.g., [NSString method:5])
3) Send to a superclass class (e.g, [super method:5] in class method)
4) Send to a superclass instance (e.g., [super method:5] in instance method)
Previously these four cases where tangled together. Now, they have
more distinct representations. Specific changes:
1) Unchanged; the object instance is represented by an Expr*.
2) Previously stored the ObjCInterfaceDecl* referring to the class
receiving the message. Now stores a TypeSourceInfo* so that we know
how the class was spelled. This both maintains typedef information
and opens the door for more complicated C++ types (e.g., dependent
types). There was an alternative, unused representation of these
sends by naming the class via an IdentifierInfo *. In practice, we
either had an ObjCInterfaceDecl *, from which we would get the
IdentifierInfo *, or we fell into the case below...
3) Previously represented by a class message whose IdentifierInfo *
referred to "super". Sema and CodeGen would use isStr("super") to
determine if they had a send to super. Now represented as a
"class super" send, where we have both the location of the "super"
keyword and the ObjCInterfaceDecl* of the superclass we're
targetting (statically).
4) Previously represented by an instance message whose receiver is a
an ObjCSuperExpr, which Sema and CodeGen would check for via
isa<ObjCSuperExpr>(). Now represented as an "instance super" send,
where we have both the location of the "super" keyword and the
ObjCInterfaceDecl* of the superclass we're targetting
(statically). Note that ObjCSuperExpr only has one remaining use in
the AST, which is for "super.prop" references.
The new representation of ObjCMessageExpr is 2 pointers smaller than
the old one, since it combines more storage. It also eliminates a leak
when we loaded message-send expressions from a precompiled header. The
representation also feels much cleaner to me; comments welcome!
This patch attempts to maintain the same semantics we previously had
with Objective-C message sends. In several places, there are massive
changes that boil down to simply replacing a nested-if structure such
as:
if (message has a receiver expression) {
// instance message
if (isa<ObjCSuperExpr>(...)) {
// send to super
} else {
// send to an object
}
} else {
// class message
if (name->isStr("super")) {
// class send to super
} else {
// send to class
}
}
with a switch
switch (E->getReceiverKind()) {
case ObjCMessageExpr::SuperInstance: ...
case ObjCMessageExpr::Instance: ...
case ObjCMessageExpr::SuperClass: ...
case ObjCMessageExpr::Class:...
}
There are quite a few places (particularly in the checkers) where
send-to-super is effectively ignored. I've placed FIXMEs in most of
them, and attempted to address send-to-super in a reasonable way. This
could use some review.
Chris Lattner [Tue, 20 Apr 2010 20:35:58 +0000 (20:35 +0000)]
push some source location information down through the compiler,
into ContentCache::getBuffer. This allows it to produce
diagnostics on the broken #include line instead of without a
location.
Daniel Dunbar [Tue, 20 Apr 2010 17:52:30 +0000 (17:52 +0000)]
IRgen: Always use i8 arrays to access union bit-fields. This is ugly, but
matches how we currently handle structs, and this correctly handles
-fno-bitfield-type-align.
Anders Carlsson [Tue, 20 Apr 2010 16:22:16 +0000 (16:22 +0000)]
Fix a bug which triggered the assertion I added yesterday. Basically, when we initialize the vtable pointer for a virtual base, and there was another path from the most derived class to another base with the same class type, we would use the wrong base.
Douglas Gregor [Tue, 20 Apr 2010 07:18:24 +0000 (07:18 +0000)]
Introduce a limit on the depth of the template instantiation backtrace
we will print with each error that occurs during template
instantiation. When the backtrace is longer than that, we will print
N/2 of the innermost backtrace entries and N/2 of the outermost
backtrace entries, then skip the middle entries with a note such as:
note: suppressed 2 template instantiation contexts; use
-ftemplate-backtrace-limit=N to change the number of template
instantiation entries shown
This should eliminate some excessively long backtraces that aren't
providing any value.
Chris Lattner [Tue, 20 Apr 2010 05:44:43 +0000 (05:44 +0000)]
don't slap noalias attribute on stret result arguments.
This mirror's Dan's patch for llvm-gcc in r97989, and
fixes the miscompilation in PR6525. There is some contention
over whether this is the right thing to do, but it is the
conservative answer and demonstrably fixes a miscompilation.
Chris Lattner [Tue, 20 Apr 2010 04:31:55 +0000 (04:31 +0000)]
revert r101863, whcih is causing Sema/altivec-init.c to fail on a ton
of buildbots with:
error: 'error' diagnostics expected but not seen:
Line 9: too few elements in vector initialization (expected 8 elements, have 2)
1 warning and 1 error generated.
Douglas Gregor [Mon, 19 Apr 2010 22:54:31 +0000 (22:54 +0000)]
Keep track of the actual storage specifier written on a variable or
function declaration, since it may end up being changed (e.g.,
"extern" can become "static" if a prior declaration was static). Patch
by Enea Zaffanella and Paolo Bolzoni.
Douglas Gregor [Mon, 19 Apr 2010 20:09:36 +0000 (20:09 +0000)]
When normal name lookup to disambiguiate an Objective-C message send
fails to find anything, perform ivar lookup and, if we find one,
consider this an instance message.
Douglas Gregor [Mon, 19 Apr 2010 18:02:19 +0000 (18:02 +0000)]
When searching for code-completion and typo-correction candidates,
look from an Objective-C class or category to its implementation, to
pick up synthesized ivars. Fixes a problem reported by David
Chisnall.
Douglas Gregor [Sun, 18 Apr 2010 18:11:38 +0000 (18:11 +0000)]
Make sure that we don't visit redeclarations of nested classes while
instantiating class members as part of an explicit
instantiation. Addresses a compilation problem in
Boost.Serialization.
Douglas Gregor [Sun, 18 Apr 2010 17:37:40 +0000 (17:37 +0000)]
C++ [namespace.memdef]p3 only applies when the friend is not named via
a qualified name. We weren't checking for an empty
nested-name-specifier when dealing with friend class templates
(although we were checking in the other places where we deal with this
paragraph). Fixes a Boost.Serialization showstopper.
Benjamin Kramer [Sun, 18 Apr 2010 12:05:54 +0000 (12:05 +0000)]
Bail out early to avoid comparing the internals of two conversion sequences of
different kinds (aka garbage). This happens if we're comparing a standard
conversion sequence to an ambiguous one which have the same KindRank.
Douglas Gregor [Sun, 18 Apr 2010 09:22:00 +0000 (09:22 +0000)]
When performing reference initialization for the purposes of overload
resolution ([over.ics.ref]), we take some shortcuts required by the
standard that effectively permit binding of a const volatile reference
to an rvalue. We have to treat lightly here to avoid infinite
recursion.
Fix the access checking of function and function template argument types,
return types, and default arguments. This fixes PR6855 along with several
similar cases where we rejected valid code.
Douglas Gregor [Sun, 18 Apr 2010 07:57:34 +0000 (07:57 +0000)]
When checking the copy constructor for the optional copy during a
reference binding to an rvalue of reference-compatible type, check
parameters after the first for complete parameter types and build any
required default function arguments. We're effectively simulating the
type-checking for a call without building the call itself.
Douglas Gregor [Sun, 18 Apr 2010 07:40:54 +0000 (07:40 +0000)]
In C++98/03, when binding a reference to an rvalue of
reference-compatible type, the implementation is permitted to make a
copy of the rvalue (or many such copies, even). However, even though
we don't make that copy, we are required to check for the presence of
a suitable copy constructor. With this change, we do.
Note that in C++0x we are not allowed to make these copies, so we test
both dialects separately.
Also note the FIXME in one of the C++03 tests, where we are not
instantiating default function arguments for the copy constructor we
pick (but do not call). The fix is obvious; eliminating the infinite
recursion it causes is not. Will address that next.
Douglas Gregor [Sun, 18 Apr 2010 02:16:12 +0000 (02:16 +0000)]
Do not consider explicit constructors when performing a copy to a
temporary object. This is blindingly obvious from reading C++
[over.match.ctor]p1, but somehow I'd missed it and it took DR152 to
educate me. Adjust one test that was relying on this non-standard
behavior.
Anders Carlsson [Sat, 17 Apr 2010 22:54:57 +0000 (22:54 +0000)]
Simplify wide bit-field layout in CGRecordLayoutBuilder, and also fix a bug where assigning to a bit-field member would overwrite other parts of the struct.
Douglas Gregor [Sat, 17 Apr 2010 22:01:05 +0000 (22:01 +0000)]
Improve our handling of user-defined conversions as part of overload
resolution. There are two sources of problems involving user-defined
conversions that this change eliminates, along with providing simpler
interfaces for checking implicit conversions:
- It eliminates a case of infinite recursion found in Boost.
- It eliminates the search for the constructor needed to copy a temporary
generated by an implicit conversion from overload
resolution. Overload resolution assumes that, if it gets a value
of the parameter's class type (or a derived class thereof), there
is a way to copy if... even if there isn't. We now model this
properly.
Add support for '-fgnu-keywords' and '-fasm' to Clang's driver. They are not
implemented precisely the same as GCC, but the distinction GCC makes isn't
useful to represent. This allows parsing code which uses GCC-specific keywords
('asm', etc.) without parsing in a fully GNU mode.
Eric Christopher [Sat, 17 Apr 2010 02:26:23 +0000 (02:26 +0000)]
Consolidate most of the integer constant expression builtin requirement
checking into a single function and use that throughout. Remove some
now unnecessary diagnostics and update tests with now more accurate
diagnostics.
projects / clang / log