Douglas Gregor [Fri, 25 Sep 2009 04:25:58 +0000 (04:25 +0000)]
Fix checking for a null pointer constant when the expression itself is
value-dependent. Audit (and fixed) all calls to
Expr::isNullPointerConstant() to provide the correct behavior with
value-dependent expressions. Fixes PR5041 and a crash in libstdc++
<locale>.
In the same vein, properly compute value- and type-dependence for
ChooseExpr. Fixes PR4996.
Anders Carlsson [Fri, 25 Sep 2009 01:54:38 +0000 (01:54 +0000)]
Who would have thought that empty classes were so tricky? Handle cases where an empty virtual base class needs to be moved aside because it conflicts with the first field.
Fix <rdar://problem/7249327> by allowing silent conversions between signed and unsigned integer values for symbolic values. This is an intermediate solution (i.e. hack) until we support extension/truncation of symbolic integers.
Douglas Gregor [Thu, 24 Sep 2009 23:14:47 +0000 (23:14 +0000)]
WIP implementation of explicit function template specialization. This
first implementation recognizes when a function declaration is an
explicit function template specialization (based on the presence of a
template<> header), performs template argument deduction + ambiguity
resolution to determine which template is being specialized, and hooks
There are many caveats here:
- We completely and totally drop any explicitly-specified template
arguments on the floor
- We don't diagnose any of the extra semantic things that we should
diagnose.
- I haven't looked to see that we're getting the right linkage for
explicit specializations
On a happy note, this silences a bunch of errors that show up in
libstdc++'s <iostream>, although Clang still can't get through the
entire header.
John McCall [Thu, 24 Sep 2009 19:53:00 +0000 (19:53 +0000)]
Refactor the representation of qualifiers to bring ExtQualType out of the
Type hierarchy. Demote 'volatile' to extended-qualifier status. Audit our
use of qualifiers and fix a few places that weren't dealing with qualifiers
quite right; many more remain.
When building CFGs, no longer reverse the statements in the CFGBlock. Instead
have the iterators and operator[] handle the traversal of statements, as they
are stored in reverse order. Tests show this has no real performance impact, but
it does simply the CFG construction logic and will make it slightly easier to
change the allocation strategy for CFGBlocks (as we have fewer copies).
Daniel Dunbar [Thu, 24 Sep 2009 05:12:36 +0000 (05:12 +0000)]
Darwin/x86-32: Enumerated types and block pointer types in structures were not
handled correctly.
- <rdar://problem/7247671> Function arguments incorrect when function returns a
struct on i386 w/ llvm-g++ and clang
Fix: <rdar://problem/7249340> [RegionStore] model stores to symbolic parameter regions
The issue was a discrepancy between how RegionStoreManager::Bind() and
RegionStoreManager::Retrieve() derived the "key" for the first element
of a symbolic region.
Douglas Gregor [Wed, 23 Sep 2009 22:26:46 +0000 (22:26 +0000)]
For code completion, note that injected-class-names found as part of
lookup in a member access expression always start a
nested-name-specifier. Additionally, rank names that start
nested-name-specifiers after other names.
This patch addresses a few issues related to 8.5.3 [dcl.init.ref]
It uses a recent API to find inherited conversion functions to do
the initializer to reference lvalue conversion (and removes a FIXME).
It issues the ambiguity diagnostics when multiple conversions are found.
WIP.
Douglas Gregor [Wed, 23 Sep 2009 14:56:09 +0000 (14:56 +0000)]
When code-completion after a "," is building an overload set, note
that there is one more argument (the one following the comma) and make
the candidate non-viable if the function cannot accept any argument in
that position.
The test/ directory needs to be created always when building objdir != srcdir,
because the toplevel 'make check-all' needs the rule for test/lit.site.cfg,
and for that test/ must exist.
Douglas Gregor [Wed, 23 Sep 2009 00:34:09 +0000 (00:34 +0000)]
Print the results of code-completion for overloading by displaying the
signature of the function with the current parameter highlighted as a
placeholder.
Douglas Gregor [Wed, 23 Sep 2009 00:16:58 +0000 (00:16 +0000)]
Separate the code-completion results for call completion from the
results for other, textual completion. For call completion, we now
produce enough information to show the function call argument that we
are currently on.
Douglas Gregor [Tue, 22 Sep 2009 23:22:24 +0000 (23:22 +0000)]
When code-completion finds a declaration only because it is usable as
the start of a nested-name-specifier, add the "::" after the
nested-name-specifier to the code-completion string.
Douglas Gregor [Tue, 22 Sep 2009 23:15:58 +0000 (23:15 +0000)]
Tweak the code-completion results ranking and formation, so that
members found in base classes have the same ranking as members found
in derived classes. However, we will introduce an informative note for
members found in base classes, showing (as a nested-name-specifier)
the qualification to name the base class, to make it clear which
members are from bases.
Douglas Gregor [Tue, 22 Sep 2009 21:41:40 +0000 (21:41 +0000)]
In C++, a variadic function does not need an ellipsis prior to the comma. Parse it in both C and C++, but diagnose it as an error in C with a fix-it hint to add the comma.
Douglas Gregor [Tue, 22 Sep 2009 21:11:38 +0000 (21:11 +0000)]
Replace the -code-completion-dump option with
-code-completion-at=filename:line:column
which performs code completion at the specified location by truncating
the file at that position and enabling code completion. This approach
makes it possible to run multiple tests from a single test file, and
gives a more natural command-line interface.
Fix a regression in accessing class getter using the dot-syntax
notation. There is still an issue accessing field of a 'Class''s isa
in legacy code using dot field access notation (as noted in the test case)
but unrelated to this patch.
Douglas Gregor [Tue, 22 Sep 2009 15:41:20 +0000 (15:41 +0000)]
Implement code completion within a function call, triggered after the
opening parentheses and after each comma. We gather the set of visible
overloaded functions, perform "partial" overloading based on the set
of arguments that we have thus far, and return the still-viable
results sorted by the likelihood that they will be the best candidate.
Most of the changes in this patch are a refactoring of the overloading
routines for a function call, since we needed to separate out the
notion of building an overload set (common to code-completion and
normal semantic analysis) and then what to do with that overload
set. As part of this change, I've pushed explicit template arguments
into a few more subroutines.
There is still much more work to do in this area. Function templates
won't be handled well (unless we happen to deduce all of the template
arguments before we hit the completion point), nor will overloaded
function-call operators or calls to member functions.