Fix up some of the visiting for array types -- we weren't good about
getting array indices before -- and for some of the builtin operators:
sizeof, offsetof, unaryops like __is_enum.
Also fix the function visitor to visit exception types in function
parameters.
Douglas Gregor [Fri, 16 Jul 2010 16:54:17 +0000 (16:54 +0000)]
When performing template name lookup for a dependent member access
expression such as the "foo" in "this->blah.foo<1, 2>", and we can't
look into the type of "this->blah" (e.g., because it is dependent),
look into the local scope of a template of the same name. Fixes
<rdar://problem/8198511>.
Douglas Gregor [Fri, 16 Jul 2010 15:40:40 +0000 (15:40 +0000)]
Revert Microsoft-specific override of the "typedef requires a name"
diagnostic. Instead, put it and the "declaration does not declare
anything" warning into -Wmissing-declarations.
Add most of the boilerplate support for scanf format string checking. This includes
handling the parsing of scanf format strings and hooking the checking into Sema.
Most of this checking logic piggybacks on what was already there for checking printf format
strings, but the checking logic has been refactored to support both.
What is left to be done is to support argument type checking in format strings and of course
fix the usual tail of bugs that will follow.
Add builtin definition for scanf, including extending the builtin encoding to
represent builtins that have the "scanf" attribution (via the format attribute) just
like we do with printf functions. Follow-up work is needed to add similar support
for fscanf et al.
This is to support format-string checking for scanf functions.
Daniel Dunbar [Fri, 16 Jul 2010 00:31:23 +0000 (00:31 +0000)]
Builtins/ARM: __clear_cache doesn't seem to have a consistent prototype, declare
the builtin as void __clear_cache(...) to workaround this, which appears to
match what GCC does.
Daniel Dunbar [Fri, 16 Jul 2010 00:00:19 +0000 (00:00 +0000)]
IRgen: Support user defined attributes on block runtime functions.
- This issue here is that /usr/include/Blocks.h wants to define some of the
block runtime globals as weak, depending on the target. This doesn't work in
Clang because we aren't using the AST decl for these globals.
- The fix is a pretty gross hack which just watches all the decls for the
specific blocks globals we need to know about; if we see one we use it,
otherwise we use the hand coded type.
In time, I would like to clean this up by changing IRgen to ask Sema/AST for
the decl, which would then be lazily loaded from the builtin table if
necessary. This could be used in a whole host of places in IRgen and would
get rid of a lot of grotty hand coding of LLVM IR; however, we need some
extra Sema support for this as well as support for builtin global variables.
John McCall [Thu, 15 Jul 2010 23:40:35 +0000 (23:40 +0000)]
When deferring the emission of declarations with initializers in C++, remember
the order they appeared in the translation unit. If they get emitted, put them
in their proper order. Fixes rdar://problem/7458115
Douglas Gregor [Thu, 15 Jul 2010 21:05:01 +0000 (21:05 +0000)]
When we're performing tentative parsing to determine whether the
parser is looking at a declaration or an expression, use a '=' to
conclude that we are parsing a declaration.
This is wrong. However, our previous approach of finding a comma after
the '=' is also wrong, because the ',' could be part of a
template-argument-list. So, for now we're going to use the same wrong
heuristic as GCC and Visual C++, because less real-world code is
likely to be broken this way. I've opened PR7655 to keep track of our
wrongness; note also the XFAIL'd test.
Douglas Gregor [Thu, 15 Jul 2010 18:58:16 +0000 (18:58 +0000)]
Reinstate the scalar-cast-to-const-reference improvements, this time
with the proper spelling of "non-class prvalue". Silly me, I think
class rvalues were xvalues rather than prvalues!
Revert 108220 and subsequent patch.
This is not required (I am not 100% sure why) but method.exp from gdb testsuite flagged regression due to this patch.
Douglas Gregor [Thu, 15 Jul 2010 18:04:13 +0000 (18:04 +0000)]
Teach CodeGenFunction::EmitCastLValue() to handle casts to an lvalue
that involve binding a reference to a pure rvalue temporary (e.g., not
a class temporary), by creating a new temporary and copying the result
there. Fixes PR6024.
Daniel Dunbar [Thu, 15 Jul 2010 15:26:14 +0000 (15:26 +0000)]
Driver/Darwin: Allow -m{ios,macosx}-version-min= to be passed with -Xarch.
- This should be safe, because the driver itself shouldn't need to make
decisions that depend on the deployment target.
Sebastian Redl [Wed, 14 Jul 2010 23:45:08 +0000 (23:45 +0000)]
Add a callback interface that allows interested parties to get notified whenever PCHReader deserializes a type or decl (and possibly other things in the future). Have PCHWriter implement these callbacks as noops and register to receive them if we're chaining PCHs. This will allow PCHWriter to track the IDs of these things, which it needs to write the dependent files. WIP
Daniel Dunbar [Wed, 14 Jul 2010 23:39:36 +0000 (23:39 +0000)]
CodeGen/ObjC/NeXT: Fix Obj-C message send to match llvm-gcc when choosing
whether to use objc_msgSend_fpret; the choice is target dependent, not Obj-C ABI
dependent.
- <rdar://problem/8139758> arm objc _objc_msgSend_fpret bug
Douglas Gregor [Wed, 14 Jul 2010 23:20:53 +0000 (23:20 +0000)]
When determining whether an overload set with explicit template
arguments only resolves to a single specialization, make sure to look
through using declarations. Fixes PR7641.
Douglas Gregor [Wed, 14 Jul 2010 23:14:12 +0000 (23:14 +0000)]
When there are extra or missing template parameter lists in a template
definition, we're likely going to end up breaking the invariants of
the template system, e.g., that the depths of template parameter lists
match up with the nesting template of the template. So, make sure we
mark such ill-formed declarations as invalid or don't even build them
at all.
Daniel Dunbar [Wed, 14 Jul 2010 18:46:27 +0000 (18:46 +0000)]
Driver: When re'execing clang, use path to the main executable instead of
looking up Clang in the normal search paths (which may end up finding the wrong
clang).
Wire up '-Wignored-qualifiers' to the warning on 'const' in 'const int f()'.
This flag and warning match GCC semantics. Also, move it to -Wextra as this is
a largely cosmetic issue and doesn't seem to mask problems. Subsequent fixes to
the tests which no longer by default emit the warning. Added explicit test
cases for both C and C++ behavior with the warning turned on.
John McCall [Wed, 14 Jul 2010 04:38:21 +0000 (04:38 +0000)]
Remove a few mangling FIXMEs:
- TSTs whose template is a template template parameter already work
- we don't provide an imaginary type, so we can't mangle one
- we don't need a generic FIXME for vendor type qualifiers
John McCall [Wed, 14 Jul 2010 04:20:34 +0000 (04:20 +0000)]
Implement the standard mangling for array-subscript expressions, and implement
the current proposals from David Vandervoorde for new, delete, throw, typeid,
imaginary literals, string literals, and null literals.
John McCall [Tue, 13 Jul 2010 22:12:14 +0000 (22:12 +0000)]
Allow for the possibility that __cxa_end_catch might throw for a catch-all block
or a catch of a record type by value or reference. Also convert this to a
lazy cleanup.
John McCall [Tue, 13 Jul 2010 20:32:21 +0000 (20:32 +0000)]
Teach IR generation how to lazily emit cleanups. This has a lot of advantages,
mostly in avoiding unnecessary work at compile time but also in producing more
sensible block orderings.
Move the destructor cleanups for local variables over to use lazy cleanups.
Eventually all cleanups will do this; for now we have some awkward code
duplication.
Tell IR generation just to never produce landing pads in -fno-exceptions.
This is a much more comprehensive solution to a problem which previously was
half-solved by checks in most cleanup-generation spots.
Chris Lattner [Tue, 13 Jul 2010 19:41:32 +0000 (19:41 +0000)]
Add a warning to catch a bug recently caught by code review, like this:
t2.c:2:12: warning: use of logical && with constant operand; switch to bitwise &
or remove constant [-Wlogical-bitwise-confusion]
return x && 4;
^ ~
Douglas Gregor [Tue, 13 Jul 2010 18:40:04 +0000 (18:40 +0000)]
Whenever we're creating an expression that is typically an rvalue
(e.g., a call, cast, etc.), immediately adjust the expression's type
to strip cv-qualifiers off of all non-class types (in C++) or all
types (in C). This effectively extends my previous fix for PR7463,
which was restricted to calls, to other kinds of expressions within
similar characteristics. I've audited every use of
getNonReferenceType() in the code base, switching to the newly-renamed
getNonLValueExprType() where necessary.
Big thanks to Eli for pointing out just how incomplete my original fix
for PR7463 actually was. We've been handling cv-qualifiers on rvalues
wrong for a very, very long time. Fixes PR7463.
Douglas Gregor [Tue, 13 Jul 2010 15:54:32 +0000 (15:54 +0000)]
Downgrade the "when type is in parentheses, array cannot have dynamic
size" error for code like
new (int [size])
to a warning, add a Fix-It to remove the parentheses, and make this
diagnostic work properly when it occurs in a template
instantiation. <rdar://problem/8018245>.
Modify the pragma handlers to accept and use StringRefs instead of IdentifierInfos.
When loading the PCH, IdentifierInfos that are associated with pragmas cause declarations that use these identifiers to be deserialized (e.g. the "clang" pragma causes the "clang" namespace to be loaded).
We can avoid this if we just use StringRefs for the pragmas.
As a bonus, since we don't have to create and pass IdentifierInfos, the pragma interfaces get a bit more simplified.
Douglas Gregor [Tue, 13 Jul 2010 08:50:30 +0000 (08:50 +0000)]
Improve diagnostics for the "type qualifier on return type has no
effect warning" by printing the qualifiers we saw and correctly
pluralizing the message, e.g.,
test/SemaCXX/conditional-expr.cpp:295:3: warning: 'const volatile' type
qualifiers on return type have no effect
const volatile Enum g2() {
^~~~~ ~~~~~~~~
Douglas Gregor [Tue, 13 Jul 2010 08:37:11 +0000 (08:37 +0000)]
When computing the canonical profile of a DeclRefExpr or MemberExpr,
don't include the nested-name-specifier or template arguments: they
were only relevant when resolving the declaration. Fixes PR7460.
Douglas Gregor [Tue, 13 Jul 2010 08:18:22 +0000 (08:18 +0000)]
When forming a function call or message send expression, be sure to
strip cv-qualifiers from the expression's type when the language calls
for it: in C, that's all the time, while C++ only does it for
non-class types.
Centralized the computation of the call expression type in
QualType::getCallResultType() and some helper functions in other nodes
(FunctionDecl, ObjCMethodDecl, FunctionType), and updated all relevant
callers of getResultType() to getCallResultType().
Fixes PR7598 and PR7463, along with a bunch of getResultType() call
sites that weren't stripping references off the result type (nothing
stripped cv-qualifiers properly before this change).
Douglas Gregor [Tue, 13 Jul 2010 06:02:28 +0000 (06:02 +0000)]
Reinstate the optimization suppressing available_externally functions
at -O0. The only change from the previous patch is that we don't try
to generate virtual method thunks for an available_externally
function.
Douglas Gregor [Tue, 13 Jul 2010 00:10:04 +0000 (00:10 +0000)]
Provide a special diagnostic for attempts to explicitly specialize
class templates within class scope (which is ill-formed), and recover
by dropping the explicit specialization entirely. Fixes the infinite
loop in PR7622.
projects / clang / log