John McCall [Tue, 11 Aug 2009 06:59:38 +0000 (06:59 +0000)]
Argument-dependent lookup for friend declarations. Add a new decl type,
FriendFunctionDecl, and create instances as appropriate.
The design of FriendFunctionDecl is still somewhat up in the air; you can
befriend arbitrary types of functions --- methods, constructors, etc. ---
and it's not clear that this representation captures that very well.
We'll have a better picture when we start consuming this data in access
control.
Douglas Gregor [Tue, 11 Aug 2009 05:31:07 +0000 (05:31 +0000)]
Refactor the template-instantiation logic for expressions into a
generic tree transformation (also used for recanonicalization) and a
small amount of template-instantiation-specific logic.
Mike Stump [Tue, 11 Aug 2009 04:03:59 +0000 (04:03 +0000)]
Implement more of the inductive case for vtable layout involving
virtual base primaries and improve the layout of classes with virtual
bases. WIP.
Hey, I've decided I want a change to FileCheck, I need to ensure that
the group is together, nothing in between. Can we change it to check
the match line is from the line immediately following the last matched
line, if the source for the matched line is immediately after the
source for the previously matched line?
// CHECK: 1
// CHECK: 2
// CHECK: 3
// CHECK: 4
// CHECK: 5
// CHECK: 6
would require 1 2 and 3 to be continuous in the output, and 4 5 and 6
to be continuous.
Nate Begeman [Mon, 10 Aug 2009 23:49:36 +0000 (23:49 +0000)]
Take 2 on AltiVec-style vector initializers.
Fixes PR4704 problems
Addresses Eli's patch feedback re: ugly cast code
Updates all postfix operators to remove ParenListExprs. While this is awful,
no better solution (say, in the parser) is obvious to me. Better solutions
welcome.
Anders Carlsson [Mon, 10 Aug 2009 21:30:22 +0000 (21:30 +0000)]
Add a CastInfo struct that will be used for cast information when constructing cast expressions. Right now it only stores the cast kind, but in the future it might store conversion functions and constructors.
Chris Lattner [Mon, 10 Aug 2009 19:03:04 +0000 (19:03 +0000)]
fix a couple of problems with section attributes:
1. Passing something that isn't a string used to cause:
"argument to annotate attribute was not a string literal"
make it say "section attribute" instead.
2. Fix the location of the above message to point to the
bad argument instead of the section token.
3. Implement rdar://4341926, by diagnosing invalid section
specifiers in the frontend rather than letting them slip all
the way to the assembler (a QoI win).
An example of #3 is that we used to produce something like this:
/var/folders/n7/n7Yno9ihEm894640nJdSQU+++TI/-Tmp-//ccFPFGtT.s:2:Expected comma after segment-name
/var/folders/n7/n7Yno9ihEm894640nJdSQU+++TI/-Tmp-//ccFPFGtT.s:2:Rest of line ignored. 1st junk character valued 46 (.).
Daniel improved clang to use llvm_report_error, so now we got:
$ clang t.c -c
fatal error: error in backend: Global variable 'x' has an invalid section specifier 'sadf': mach-o section specifier
requires a segment and section separated by a comma.
with no loc info. Now we get:
$ clang t.c -fsyntax-only
t.c:4:30: error: argument to 'section' attribute is not valid for this target: mach-o section specifier requires a segment
and section separated by a comma
int x __attribute__((section("sadf")));
^
Daniel Dunbar [Mon, 10 Aug 2009 03:40:28 +0000 (03:40 +0000)]
Use LLVM's new error handler API to report back end errors using Diagnostic.
For example,
--
ddunbar@giles:Frontend$ clang -c backend-errors.c
fatal error: error in backend: Global variable 'a' has an invalid section specifier
'I AM, not, legal': mach-o section specifier uses an unknown section type.
--
compare to:
--
ddunbar@giles:Frontend$ gcc -c backend-errors.c
/var/folders/DQ/DQ8GT3++HESEzT1obWBynE+++TI/-Tmp-//cc45w2pq.s:2:Expected comma after segment-name
/var/folders/DQ/DQ8GT3++HESEzT1obWBynE+++TI/-Tmp-//cc45w2pq.s:2:Rest of line ignored. 1st junk character valued 77 (M).
--
Yay!
I am not tied to my wording choice, we could also go with "uncoverable error"
for the prefix, or just leave it off entirely.
Daniel Dunbar [Mon, 10 Aug 2009 03:01:36 +0000 (03:01 +0000)]
Revert r78535, it is causing a number of failures to build projects.
--- Reverse-merging r78535 into '.':
D test/Sema/altivec-init.c
U include/clang/Basic/DiagnosticSemaKinds.td
U include/clang/AST/Expr.h
U include/clang/AST/StmtNodes.def
U include/clang/Parse/Parser.h
U include/clang/Parse/Action.h
U tools/clang-cc/clang-cc.cpp
U lib/Frontend/PrintParserCallbacks.cpp
U lib/CodeGen/CGExprScalar.cpp
U lib/Sema/SemaInit.cpp
U lib/Sema/Sema.h
U lib/Sema/SemaExpr.cpp
U lib/Sema/SemaTemplateInstantiateExpr.cpp
U lib/AST/StmtProfile.cpp
U lib/AST/Expr.cpp
U lib/AST/StmtPrinter.cpp
U lib/Parse/ParseExpr.cpp
U lib/Parse/ParseExprCXX.cpp
In addition to being defined by the AltiVec PIM, this is also the vector
initializer syntax used by OpenCL, so that vector literals are compatible
with macro arguments.
Anders Carlsson [Sat, 8 Aug 2009 18:23:56 +0000 (18:23 +0000)]
Introduce a new PragmaPack attribute, and use it for #pragma pack. The PackedAttr now only represents __attribute__((packed)).
This is necessary because #pragma pack and __attribute__((packed)) have different semantics. No functionality change yet, but this lays the groundwork for fixing a record layout bug.
Douglas Gregor [Sat, 8 Aug 2009 01:41:12 +0000 (01:41 +0000)]
Introduce reference counting for statements and expressions, using it
to allow sharing of nodes. Simplifies some aspects of template
instantiation, and fixes both PR3444 and <rdar://problem/6757457>.
John McCall [Fri, 7 Aug 2009 22:18:02 +0000 (22:18 +0000)]
Just add global scope to the associated namespaces set instead of tracking it
separately. Add the framework (currently unfed) for finding friend declarations
during argument-dependent lookup.
Douglas Gregor [Fri, 7 Aug 2009 06:08:38 +0000 (06:08 +0000)]
Separate Stmt::Destroy into the entrypoint for destroying a statement
or expression (Destroy) from the virtual function used to actually
destroy a given expression (DoDestroy).
Ted Kremenek [Thu, 6 Aug 2009 22:33:36 +0000 (22:33 +0000)]
Fix a few more false positives involving RegionStore and unions, but this time
with array accesses. In the process, refactor some common logic in
RetrieveElement() and RetrieveField() into RetrieveFieldOrElementCommon().
John McCall [Thu, 6 Aug 2009 20:49:32 +0000 (20:49 +0000)]
Permit a class to friend its members in C++0x, without restriction.
Permit a class to friend its class members in C++ 98, as long as extensions
are enabled (and even when they aren't, only give an extwarn about it).
Douglas Gregor [Thu, 6 Aug 2009 16:20:37 +0000 (16:20 +0000)]
When we encounter a dependent type that was parsed before we know that
we were going to enter into the scope of a class template or class
template partial specialization, rebuild that type so that it can
refer to members of the current instantiation, as in code like
Without rebuilding the return type of this out-of-line definition, the
canonical return type of the out-of-line definition (a TypenameType)
will not match the canonical return type of the declaration (the
canonical type of T*).
Mike Stump [Thu, 6 Aug 2009 14:12:47 +0000 (14:12 +0000)]
Testcase for a recent checkin for vbase layout ordering. I hate to
optimize and check for non-optimal code, but until the frontend is as
powerful as fold...
projects / clang / log