Douglas Gregor [Tue, 7 Jul 2015 03:57:15 +0000 (03:57 +0000)]
Parsing, semantic analysis, and AST for Objective-C type parameters.
Produce type parameter declarations for Objective-C type parameters,
and attach lists of type parameters to Objective-C classes,
categories, forward declarations, and extensions as
appropriate. Perform semantic analysis of type bounds for type
parameters, both in isolation and across classes/categories/extensions
to ensure consistency.
Also handle (de-)serialization of Objective-C type parameter lists,
along with sundry other things one must do to add a new declaration to
Clang.
Note that Objective-C type parameters are typedef name declarations,
like typedefs and C++11 type aliases, in support of type erasure.
Adrian Prantl [Tue, 7 Jul 2015 00:49:35 +0000 (00:49 +0000)]
Debug info: Emit distinct __block_literal_generic types for blocks with
different function signatures. (Previously clang would emit all block
pointer types with the type of the first block pointer in the compile
unit.)
This reverts commit r241244, but restricts SEH support to Win64.
This way, Chromium builds will still fall back on TUs with SEH, and
Clang developers can work on this incrementally upstream while patching
this small predicate locally. It'll also make it easier to review small
fixes.
Davide Italiano [Mon, 6 Jul 2015 18:02:09 +0000 (18:02 +0000)]
[Sema] Warn when shifting a negative value.
Example:
% ./clang -Wshift-negative-value emit.c
emit.c:3:14: warning: shifting a negative signed value is undefined [-Wshift-negative-value]
int a = -1 << 3;
~~ ^
1 warning generated.
Teresa Johnson [Mon, 6 Jul 2015 16:23:00 +0000 (16:23 +0000)]
Resubmit "Pass down the -flto option to the -cc1 job" (r239481)
The patch is the same except for the addition of a new test for the
issue that required reverting the dependent llvm commit.
--Original Commit Message--
Pass down the -flto option to the -cc1 job, and from there into the
CodeGenOptions and onto the PassManagerBuilder. This enables gating
the new EliminateAvailableExternally module pass on whether we are
preparing for LTO.
If we are preparing for LTO (e.g. a -flto -c compile), the new pass is not
included as we want to preserve available externally functions for possible
link time inlining.
Daniel Jasper [Mon, 6 Jul 2015 11:30:34 +0000 (11:30 +0000)]
clang-format: Fix __attribute__ being treated as decl name.
__attribute__ was treated as the name of a function definition, with the
tokens in parentheses being the parameter list. This formats incorrectly
with AlwaysBreakAfterDefinitionReturnType. Fix it by treating
__attribute__ like decltype.
Richard Smith [Mon, 6 Jul 2015 04:43:58 +0000 (04:43 +0000)]
PR24030, PR24033: Consistently check whether a new declaration conflicts with
an existing using shadow declaration if they define entities of the same kind
in different namespaces.
We'd previously check this consistently if the using-declaration came after the
other declaration, but not if it came before.
[CMake] clang-*.*: Prevent versioning if the buildhost is targeting for Win32.
CMake-2.8.12 is hardcoded to create symlinked clang.exe if the target property VERSION is present and the host is not Win32.
Then clang.exe-*.* is generated and clang.exe is symlinked to it.
- Refactored ARMTargetInfo in order to use the API of llvm/lib/Support/TargetParser.cpp
for extracting target specific information.
- Patch for commit 241267: ShouldUseInlineAtomic was set incorrectly when subArch was
not specified, causing regressions.
Driver: Replace a couple of out of date terms in the docs
The Job base class was removed in r241310, so replace a couple of
references to it with Command. Also change another use of Job with
Action, since that's the term used in the source.
We had a strange relationship here where we made a list of Jobs
inherit from a single Job, but there weren't actually any places where
this arbitrary nesting was used or needed.
Simplify all of this by removing Job entirely and updating all of the
users to either work with a JobList or a single Command.
When testing for anyOf(), the test should not be for an exact type match for all members of the set. Instead, test that all members are convertible to the common type.
Ben Langmuir [Thu, 2 Jul 2015 13:19:48 +0000 (13:19 +0000)]
[Modules] Be consistent about finding a module for framework headers
We use findModuleForHeader() in several places, but in header search we
were not calling it when a framework module didn't show up with the
expected name, which would then lead to unexpected non-modular includes.
Now we will find the module unconditionally for frameworks. For regular
frameworks, we use the spelling of the module name from the module map
file, and for inferred ones we use the canonical directory name.
In the future we might want to lock down framework modules sufficiently
that these name mismatches cannot happen.
David Majnemer [Thu, 2 Jul 2015 09:43:11 +0000 (09:43 +0000)]
[MS ABI] nullptr data member ptrs are mangled differently for classes vs fns
It turns out that nullptr pointers to data members act differently in
function templates vs class templates. Class templates use a variable
width representation proportional to the number of fields needed to
materialize it. Function templates always use a single '0' template
parameter. However, using '0' all the time is problematic if the class
uses single or multiple inheritance. In those cases, use -1.
Objective-C format strings now support modifier flags
that can be attached to a '@' conversion. Currently
the only one supported, as of iOS 9 and OS X 10.11,
is the new "technical term", denoted by the flag "tt",
for example:
%[tt]@
instead of just:
%@
The 'tt' stands for "technical term", which is used
by the string-localization facilities on Darwin to
add the appropriate spacing or quotation depending
the language locale.
Support mingw-w64 and mingw.org toolchains at any install location.
No more hardcoded paths: clang will use -sysroot as gcc root location if
provided. Otherwise, it will search for gcc on the path. If not found it
will use the driver installed location.
http://reviews.llvm.org/D5268
Patch by Ruben Van Boxem, Martell Malone, Yaron Keren.
Reviewed by Reid Kleckner.
[OPENMP 4.0] Codegen for 'cancellation point' directive.
The next code is generated for this construct:
```
if (__kmpc_cancellationpoint(ident_t *loc, kmp_int32 global_tid, kmp_int32 cncl_kind) != 0)
<exit from outer innermost construct>;
```
[OPENMP] Introduced type trait "__builtin_omp_required_simd_align" for default simd alignment.
Adds type trait "__builtin_omp_required_simd_align" after discussions here http://reviews.llvm.org/D9894
Differential Revision: http://reviews.llvm.org/D10597
Driver: add support for linking the UBSan runtime library on Windows.
On Windows the user may invoke the linker directly, so we might not have an
opportunity to add runtime library flags to the linker command line. Instead,
instruct the code generator to embed linker directive in the object file
that cause the required runtime libraries to be linked.
We might also want to do something similar for ASan, but it seems to have
its own special complexities which may make this infeasible.
Richard Smith [Wed, 1 Jul 2015 23:19:58 +0000 (23:19 +0000)]
[modules] Merging support for specializations of a function template. This very
rarely matters, but can affect whether two dependent types are canonically
equivalent.
Anton Yartsev [Wed, 1 Jul 2015 22:35:29 +0000 (22:35 +0000)]
[analyzer] Prevent ccc/c++-analyzer from hanging on Windows.
'fork'+'exec' combination made scan-build and ccc-analyzer hang under Windows. The patch replaces 'fork'+'exec' with more reliable 'system' (ccc-analyzer) and piped 'open' (scan-build). See http://reviews.llvm.org/D8774 and http://reviews.llvm.org/D9357 for more details.
[SEH] Delete the 32-bit IR lowering for __finally blocks and use x64
32-bit finally funclets are intended to be called both directly from the
parent function and indirectly from the EH runtime. Because we aren't
contorting LLVM's X86 prologue to match MSVC's, calling the finally
block directly passes in a different value of EBP than the one that the
runtime provides. We need an adapter thunk to adjust EBP to the expected
value. However, WinEHPrepare already has to solve this problem when
cleanups are not pre-outlined, so we can go ahead and rely on it rather
than duplicating work.
Now we only do the llvm.x86.seh.recoverfp dance for 32-bit SEH filter
functions.
Richard Smith [Wed, 1 Jul 2015 19:32:54 +0000 (19:32 +0000)]
[modules] Don't make out-of-line member specializations of an instantiated
class template specialization visible just because the class template
specialization's definition is visible.
This re-lands r236052 and adds support for __exception_code().
In 32-bit SEH, the exception code is not available in eax. It is only
available in the filter function, and now we arrange to load it and
store it into an escaped variable in the parent frame.
As a consequence, we have to disable the "catch i8* null" optimization
on 32-bit and always generate a filter function. We can re-enable the
optimization if we detect an __except block that doesn't use the
exception code, but this probably isn't worth optimizing.
[DebugInfo] Fix debug info generation for function static variables, typedefs, and records
Function static variables, typedefs and records (class, struct or union) declared inside
a lexical scope were associated with the function as their parent scope, rather than the
lexical scope they are defined or declared in.
-frewrite-includes: Rework how includes and modules are differentiated
The map of FileChange structs here was storing two disjoint types of
information:
1. A pointer to the Module that an #include directive implicitly
imported
2. A FileID and FileType for an included file. These would be left
uninitialized in the Module case.
This change splits these two kinds of information into their own maps,
which both simplifies how we access either and avoids the undefined
behaviour we were hitting due to the uninitialized fields in the
included file case.
Mostly NFC, but fixes some errors found by self-host with ubsan.
Richard Smith [Wed, 1 Jul 2015 02:29:35 +0000 (02:29 +0000)]
[modules] Before checking whether the controlling macro of a header is defined,
update the identifier in case we've imported a definition of the macro (and
thus the contents of the header) from a module.
Also fold ExternalIdentifierLookup into ExternalPreprocessorSource; it no longer
makes sense to keep these separate now that the only user of the former also
needs the latter.