Eli Friedman [Wed, 2 Nov 2011 23:06:23 +0000 (23:06 +0000)]
Use native endianness for writing out character escapes to the result buffer for string literal parsing. No functional change on little-endian architectures; should fix test failures on PPC.
Anna Zaks [Wed, 2 Nov 2011 20:46:50 +0000 (20:46 +0000)]
[analyzer] Tweak the test script
Rename the pre_run_static_analyzer script into cleanup_run_static_analyzer
and use it to clean up after the reference build - which is done as part of
adding a new project to the system.
Douglas Gregor [Wed, 2 Nov 2011 17:38:53 +0000 (17:38 +0000)]
Drastically simplify the mapping from the declaration corresponding to
the injected-class-name of a class (or class template) to the
declaration that results from substituting the given template
arguments. Previously, we would actually perform a substitution into
the injected-class-name type and then retrieve the resulting
declaration. However, in certain, rare circumstances involving
deeply-nested member templates, we would get the wrong substitution
arguments.
This new approach just matches up the declaration with a declaration
that's part of the current context (or one of its parents), which will
either be an instantiation (during template instantiation) or the
declaration itself (during the definition of the template). This is
both more efficient (we're avoiding a substitution) and more correct
(we can't get the template arguments wrong in the member-template
case).
Bob Wilson [Wed, 2 Nov 2011 05:10:45 +0000 (05:10 +0000)]
Invoke the Darwin assembler with -g instead of --gdwarf2.
The -g and --gdwarf2 options are currently synonyms to the Darwin assembler.
But clang itself does not recognize --gdwarf2, so if we want to experiment
with using clang, with its integrated assembler, to replace the default
assembler, it is necessary to use -g. <rdar://problem/10349486>
Update the CMake build for clang-tblgen to reflect that it directly
depends on the Support library rather than relying on TableGen's
transitive dependency.
Douglas Gregor [Tue, 1 Nov 2011 22:13:30 +0000 (22:13 +0000)]
When we run into a constructor or destructor that is defined in the
wrong class, make sure to drop it immediately; we don't want that
constructor to be available within the DeclContext. Fixes
<rdar://problem/9677163>.
Douglas Gregor [Tue, 1 Nov 2011 21:35:16 +0000 (21:35 +0000)]
When we see an out-of-line definition of a member class template that
does not match any declaration in the class (or class template), be
sure to mark it as invalid. Fixes PR10924 / <rdar://problem/10119422>.
Richard Smith [Tue, 1 Nov 2011 20:38:59 +0000 (20:38 +0000)]
Temporarily disable lvalue-to-rvalue conversions on const pointers while an
apparent miscompile triggered by this is investigated. This is essentially a
revert of r143298.
[libclang] For a class extension, give it a unique USR but for any property or ivar
it contains give it a USR based on its semantic context, which is the interface.
This follows what we already did for objc methods. rdar://10371669
Eli Friedman [Tue, 1 Nov 2011 02:23:42 +0000 (02:23 +0000)]
Fix the representation of wide strings in the AST and IR so that it uses the native representation of integers for the elements. This fixes a bunch of nastiness involving
treating wide strings as a series of bytes.
Eli Friedman [Tue, 1 Nov 2011 02:14:50 +0000 (02:14 +0000)]
Perform proper conversion for strings encoded in the source file as UTF-8. (For now, we are assuming the source character set is always UTF-8; this can be easily extended if necessary.)
Tobias Grosser [Tue, 1 Nov 2011 01:34:59 +0000 (01:34 +0000)]
Add LinkAllPasses to clang
This patch ensures that no passes are deleted from clang, such that the loading
of plugins does not fail because of passes being unavailable. This increases
the size of the clang binary from 43029853 to 43915291 bytes (around 2%).
Douglas Gregor [Tue, 1 Nov 2011 01:16:03 +0000 (01:16 +0000)]
Rework the AST for the initializer of a delegating constructor, so
that it retains source location information for the type. Aside from
general goodness (being able to walk the types described in that
information), we now have a proper representation for dependent
delegating constructors. Fixes PR10457 (for real).
Adds IRGen support for captured rvalue references in blocks.
In this case, temporary value is copied into block descriptor
as their own copy to work on. // rdar://9971124
Ted Kremenek [Mon, 31 Oct 2011 22:23:51 +0000 (22:23 +0000)]
[libclang] Tweak internals of CXSourceLocation to allow an alternate implementation if the lowest bit of ptr_data[0] is not 0. This
is prep for work on serialized diagnostics.
Ted Kremenek [Mon, 31 Oct 2011 21:40:19 +0000 (21:40 +0000)]
[libclang] Add CXDiagnosticImpl to represent a super class for the implementation backing a CXDiagnostic. This allows CXStoredDiagnostic
to be just one possible implementation of a CXDiagnostic.
Revert r143342. Caching of code-completion results was intentionally placed in "reparse"
because we don't want to take this performance hit when doing code completion
Log of r143342:
Move caching of code-completion results from ASTUnit::Reparse to ASTUnit::CodeComplete,
so that it will happen when we are doing code-completion, not reparsing.
Anders Carlsson [Mon, 31 Oct 2011 16:27:11 +0000 (16:27 +0000)]
In x86_64, when calling an Objective-C method that returns a _Complex long double, make sure to use the objc_msgSend_fp2ret function which ensures that the return value will be {0, 0} if the receiver is nil.
Axel Naumann [Mon, 31 Oct 2011 11:02:24 +0000 (11:02 +0000)]
Add possibility to specify an include location for createFileIDForMemBuffer().
Useful to set up a virtual include hierarchy, e.g. to order diagnostics for ambiguous overloads from MemoryBuffers.
Chandler Carruth [Mon, 31 Oct 2011 09:06:40 +0000 (09:06 +0000)]
Simplify the set of directories we look for on multiarch systems.
I don't have any Debian system with one of these currently, and it seems
unlikely for one to show up suddenly. We can add more patterns here if
they become necessary.
Chandler Carruth [Mon, 31 Oct 2011 08:42:27 +0000 (08:42 +0000)]
Add "just one more" include path to the monstrosity that is our header
search logic. The Debian multiarch seems to have completely changed from
when these were originally added, and I'd like to remove a bunch of
them, but I'll be lazy and delay that until this logic is hoisted into
the driver where it belongs.
Chandler Carruth [Mon, 31 Oct 2011 08:42:24 +0000 (08:42 +0000)]
Fix part of PR11223 and probably a few dups as well. This teaches the
library search logic to "properly" handle multiarch installations. I've
tested this on both Debian unstable and the latest Ubuntu which both use
this setup, and this appears to work largely the same way as GCC does.
It isn't exactly the same, but it is close enough and more principled in
its behavior where it differs. This should resolve any failures to find
'crt1.o' etc on Debian-based Linux distributions. If folks find more
cases where we fail, please file bugs and CC me.
Test cases for all of the debian silliness are waiting both to simplify
the process of merging these down into the 3.0 release, and because
they're so crazy I haven't yet been able to really produce a fake tree
that represents what we need to test for. I'll eventually add them
though.
Move caching of code-completion results from ASTUnit::Reparse to ASTUnit::CodeComplete,
so that it will happen when we are doing code-completion, not reparsing.
Richard Smith [Mon, 31 Oct 2011 05:11:32 +0000 (05:11 +0000)]
C++11 generalized constant expressions: evaluate equality comparisons between
arbitrary pointers, if those pointers don't point to weak objects or literals.
Richard Smith [Sat, 29 Oct 2011 21:53:17 +0000 (21:53 +0000)]
constexpr evaluation: allow lvalue-to-rvalue conversion on any literal type, not
just integers and floating point types. Since we don't support evaluating class
types or performing lvalue-to-rvalue conversions on array elements yet, this
just means pointer types right now.
Richard Smith [Sat, 29 Oct 2011 20:57:55 +0000 (20:57 +0000)]
constexpr function substitution:
Track the function invocation where an lvalue referring to a constexpr function
parameter originated from, and use it to substitute the correct argument and to
determine whether such an argument's lifetime has ended.
Richard Smith [Sat, 29 Oct 2011 00:50:52 +0000 (00:50 +0000)]
Rename Expr::Evaluate to Expr::EvaluateAsRValue to make it clear that it will
implicitly perform an lvalue-to-rvalue conversion if used on an lvalue
expression. Also improve the documentation of Expr::Evaluate* to indicate which
of them will accept expressions with side-effects.
Ted Kremenek [Sat, 29 Oct 2011 00:12:39 +0000 (00:12 +0000)]
Start work on SerializedDiagnosticPrinter, a new DiagnosticConsumer that serializes out the diagnostics for a given translation unit to a bit code file. This is a WIP.
The motivation for this new DiagnosticConsumer is to provide a way for tools invoking the compiler
to get its diagnostics via a libclang interface, rather than textually parsing the compiler output.
This gives us flexibility to change the compiler's textual output, but have a structured data format
for clients to use to get the diagnostics via a stable API.
I have no tests for this, but llvm-bcanalyzer so far shows that the emitted file is well-formed.
projects / clang / log