attached. Since we do not support any attributes which appertain to a statement
(yet), testing of this is necessarily quite minimal.
Patch by Alexander Kornienko!
llvm-svn: 154723
jump into these scopes, and the cleanup-entering code sometimes wants
to do some operations first (e.g. a GEP), which can leave us with
unparented IR.
llvm-svn: 154684
shadow of a block expression with non-trivial destructed cleanups,
we should flag that in the enclosing function, not in the block
that we're about to pop.
llvm-svn: 154646
implementations, mark the atomics-related parts of the C++11 status page
as done. I've not marked 'Strong Compare and Exchange' done, since although
we implement supporting builtins, we don't yet produce different code for
the weak and strong forms.
llvm-svn: 154644
__atomic_test_and_set, __atomic_clear, plus a pile of undocumented __GCC_*
predefined macros.
Implement library fallback for __atomic_is_lock_free and
__c11_atomic_is_lock_free, and implement __atomic_always_lock_free.
Contrary to their documentation, GCC's __atomic_fetch_add family don't
multiply the operand by sizeof(T) when operating on a pointer type.
libstdc++ relies on this quirk. Remove this handling for all but the
__c11_atomic_fetch_add and __c11_atomic_fetch_sub builtins.
Contrary to their documentation, __atomic_test_and_set and __atomic_clear
take a first argument of type 'volatile void *', not 'void *' or 'bool *',
and __atomic_is_lock_free and __atomic_always_lock_free have an argument
of type 'const volatile void *', not 'void *'.
With this change, libstdc++4.7's <atomic> passes libc++'s atomic test suite,
except for a couple of libstdc++ bugs and some cases where libc++'s test
suite tests for properties which implementations have latitude to vary.
llvm-svn: 154640
The codepath already only works for source bits > target bits, it's just that
it was testing for the source expr bits to be exactly 64. This meant simple
cases (int i = x_long / 2) were missed & ended up under the general
-Wconversion warning, which a user might not have enabled.
llvm-svn: 154626
We should not deserialize unused declarations from the PCH file. Achieve
this by storing the top level declarations during parsing
(HandleTopLevelDecl ASTConsumer callback) and analyzing/building a call
graph only for those.
Tested the patch on a sample ObjC file that uses PCH. With the patch,
the analyzes is 17.5% faster and clang consumes 40% less memory.
Got about 10% overall build/analyzes time decrease on a large Objective
C project.
A bit of CallGraph refactoring/cleanup as well..
llvm-svn: 154625
in general (such an atomic has boolean representation) and
specifically for IR generation of __c11_atomic_init. The latter also
means actually using initialization semantics for this initialization,
rather than just creating a store.
On a related note, make sure we actually put in non-atomic-to-atomic
conversions when performing an implicit conversion sequence. IR
generation is far too kind here, but we still want the ASTs to make
sense.
llvm-svn: 154612
types. The second and third conversions in the sequence are based on
the conversion for the underlying type, so that we get sensible
overloading behavior for, e.g., _Atomic(int) vs. _Atomic(float).
As part of this, actually implement the lvalue-to-rvalue conversion
for atomic types. There is probably a pile of code in SemaExpr that
can now be deleted, but I haven't tracked it down yet.
llvm-svn: 154596
CompilerInstance::setCodeCompletionConsumer instead, in order to change
the SkipFunctionBodies flag accordingly. Also fixed
setCodeCompletionConsumer to take a reset() to null into account.
llvm-svn: 154585
turns out that it's actually needed for C++ modules support. Since simplifying
it didn't cause any test failures, I'll add a test for it.
llvm-svn: 154582
This is not quite sufficient for libstdc++'s <atomic>: we still need
__atomic_test_and_set and __atomic_clear, and may need a more complete
__atomic_is_lock_free implementation.
We are also missing an implementation of __atomic_always_lock_free,
__atomic_nand_fetch, and __atomic_fetch_nand, but those aren't needed
for libstdc++.
llvm-svn: 154579
get the diagnostic category name from a serialized diagnostic when the version of libclang used
to read the diagnostic file is newer than the clang that emitted the diagnostic file.
llvm-svn: 154567
As per Jordy's review. Creating a symbol here is more flexible; however
I could not come up with an example where it was needed. (What
constrains can be added on of the symbol constrained to 0?)
llvm-svn: 154542
to get at the parameters (and their types) of a function or objc method cursor.
int clang_Cursor_getNumArguments(CXCursor C);
CXCursor clang_Cursor_getArgument(CXCursor C, unsigned i);
rdar://11201527
llvm-svn: 154523
<stdatomic.h> header.
In passing, fix LanguageExtensions to note that C11 and C++11 are no longer
"upcoming standards" but are now actually standardized.
llvm-svn: 154513
output the errors that occurred even if we did not get an AST (e.g. because the
PCH failed to load).
Also honor displayDiagnostics in clang_indexSourceFile().
rdar://11203489
llvm-svn: 154472
(Applied changes to CStringAPI, Malloc, and Taint.)
This might almost never happen, but we should not crash even if it does.
This fixes a crash on the internal analyzer buildbot, where postgresql's
configure was redefining memmove (radar://11219852).
llvm-svn: 154451
and emit a relatively empty block for a plain break statement. This
enables us to track where we went through a switch.
PR9796 & rdar://11215207
llvm-svn: 154420
code-completion related strings specific to a translation unit (ASTContext and related data)
CodeCompletionAllocator does such limited caching, by caching the name assigned
to a DeclContext*, but that is not the appropriate place since that object has
a lifetime that can extend beyond that of an ASTContext.
Introduce CodeCompletionTUInfo which will be always tied to a translation unit
to do this kind of caching and move the caching of CodeCompletionAllocator into this
object, and propagate it to all the places where it will be needed.
The plan is to extend the caching where appropriate, using CodeCompletionTUInfo,
to avoid re-calculating code-completion strings.
Part of rdar://10796159.
llvm-svn: 154408
* Alternative tokens (such as 'compl') are treated as identifiers in
attribute names.
* An attribute-list can start with a comma.
* An ellipsis may not be used with either of our currently-supported
C++11 attributes.
llvm-svn: 154381
* In C++11, '[[' is ill-formed unless it starts an attribute-specifier. Reject
array sizes and array indexes which begin with a lambda-expression. Recover by
parsing the lambda as a lambda.
* In Objective-C++11, either '[' could be the start of a message-send.
Fully disambiguate this case: it turns out that the grammars of message-sends,
lambdas and attributes do not actually overlap. Accept any occurrence of '[['
where either '[' starts a message send, but reject a lambda in an array index
just like in C++11 mode.
Implement a couple of changes to the attribute wording which occurred after our
attributes implementation landed:
* In a function-declaration, the attributes go after the exception specification,
not after the right paren.
* A reference type can have attributes applied.
* An 'identifier' in an attribute can also be a keyword. Support for alternative
tokens (iso646 keywords) in attributes to follow.
And some bug fixes:
* Parse attributes after declarator-ids, even if they are not simple identifiers.
* Do not accept attributes after a parenthesized declarator.
* Accept attributes after an array size in a new-type-id.
* Partially disamiguate 'delete' followed by a lambda. More work is required
here for the case where the lambda-introducer is '[]'.
llvm-svn: 154369
requires the -plugin to come before any -plugin-opt options, we were passing
them the other way around. With this one can run (for example):
clang -o foo foo.c -O4 -Wl,-plugin-opt=generate-api-file
llvm-svn: 154357
base-class subojects.
Incidentally, thinking about virtual bases makes it clear to me that
we're not appropriately computing the access to the virtual base's
member because we're not computing the best possible access to the
virtual base at all; in fact, we're basically assuming it's public.
I'll file a separate PR about that.
llvm-svn: 154346
to define a special member function as deleted so that it properly
establishes an object context for the accesses to the base subobject
members.
llvm-svn: 154343
We currently want to look whether PWD is available - if PWD is available it will
get us the non-resolved current path, while fs::current_path will resolve
symlinks. The long term fix is to not rely on that behavior any more.
llvm-svn: 154330
Specifically, using a an integer outside [0, 1] as a boolean constant seems to
be an easy mistake to make with things like "x == a || b" where the author
intended "x == a || x == b".
The bug caused by calling SkipUntil with three token kinds was also identified
by a VC diagnostic & reported by Francois Pichet as review feedback for my
commit r154163. I've included test cases to verify the error recovery that was
broken/poorly implemented due to this bug.
The other fix (lib/Sema/SemaExpr.cpp) seems like that code was never actually
reached in any of Clang's tests & is related to Objective C features I'm not
familiar with, so I've not been able to construct a test case for it. Perhaps
someone else can.
llvm-svn: 154325
case as we don't currently have any way of dumping target options or
otherwise observing this. Another small step toward fixing PR12380. With
this we generate TLS accesses using the static model instead of the
dynamic model, but we're still generating suboptimal code under the
mistaken assumption that the TLS offset might be greater than 2^32, and
therefor not viable as an immediate offset of a segment register.
llvm-svn: 154298
First, this patch cleans up the parsing of the PIC and PIE family of
options in the driver. The existing logic failed to claim arguments all
over the place resulting in kludges that marked the options as unused.
Instead actually walk all of the arguments and claim them properly.
We now treat -f{,no-}{pic,PIC,pie,PIE} as a single set, accepting the
last one on the commandline. Previously there were lots of ordering bugs
that could creep in due to the nature of the parsing. Let me know if
folks would like weird things such as "-fPIE -fno-pic" to turn on PIE,
but disable full PIC. This doesn't make any sense to me, but we could in
theory support it.
Options that seem to have intentional "trump" status (-static, -mkernel,
etc) continue to do so and are commented as such.
Next, a -pie-level flag is threaded into the frontend, rigged to
a language option, and handled preprocessor, setting up the appropriate
defines. We'll now have the correct defines when compiling with -fpie.
The one place outside of the preprocessor that was inspecting the PIC
level (as opposed to the relocation model, which is set and handled
separately, yay!) is in the GNU ObjC runtime. I changed it to exactly
preserve existing behavior. If folks want to change its behavior in the
face of PIE, they can do that in a separate patch.
Essentially the only functionality changed here is the preprocessor
defines and bug-fixes to the argument management.
Tests have been updated and extended to test all of this a bit more
thoroughly.
llvm-svn: 154291
testing any of the strange driver behavior. We already have some tiny
tests for the driver behavior, and I'm going to expand them greatly in
the next commit.
llvm-svn: 154290
- The [class.protected] restriction is non-trivial for any instance
member, even if the access lacks an object (for example, if it's
a pointer-to-member constant). In this case, it is equivalent to
requiring the naming class to equal the context class.
- The [class.protected] restriction applies to accesses to constructors
and destructors. A protected constructor or destructor can only be
used to create or destroy a base subobject, as a direct result.
- Several places were dropping or misapplying object information.
The standard could really be much clearer about what the object type is
supposed to be in some of these accesses. Usually it's easy enough to
find a reasonable answer, but still, the standard makes a very confident
statement about accesses to instance members only being possible in
either pointer-to-member literals or member access expressions, which
just completely ignores concepts like constructor and destructor
calls, using declarations, unevaluated field references, etc.
llvm-svn: 154248
However, the '-x' option has special handling and wasn't following this
paradigm. Fix it to do so by claiming the arg as we parse the '-x' option.
rdar://11203340
llvm-svn: 154231
In a few cases clang emitted a rather content-free diagnostic: 'parse error'.
This change replaces two actual cases (template parameter parsing and K&R
parameter declaration parsing) with more specific diagnostics and removes a
third dead case of this in the BalancedDelimiterTracker (the ctor already
checked the invariant necessary to ensure that the diag::parse_error was never
actually used).
llvm-svn: 154224
we use the same Expr* as the one being currently visited. This is preparation for transitioning to having
ProgramPoints refer to CFGStmts.
This required a bit of trickery. We wish to keep the old Expr* bindings in the Environment intact,
as plenty of logic relies on it and there is no reason to change it, but we sometimes want the Stmt* for
the ProgramPoint to be different than the Expr* being used for bindings. This requires adding an extra
argument for some functions (e.g., evalLocation). This looks a bit strange for some clients, but
it will look a lot cleaner when were start using CFGStmt* in the appropriate places.
As some fallout, the diagnostics arrows are a bit difference, since some of the node locations have changed.
I have audited these, and they look reasonable.
llvm-svn: 154214
This method is very hot, it is called when emitting diagnostics, in -E mode
and for many #pragma handlers. It scans through the whole source file to
count newlines, records and caches them in a vector.
The speedup from vectorization isn't very large, as we fall back to bytewise
scanning when we hit a newline. There might be a way to avoid leaving the sse
loop but everything I tried didn't work out because a call to push_back
clobbers xmm registers.
About 2% speedup on average on "clang -E > /dev/null" of all .cpp files in
clang's lib/Sema.
llvm-svn: 154204
global destructor entry. For some reason this isn't enabled for
apple-kexts; it'd be good to have documentation for that.
Based on a patch by Nakamura Takumi!
llvm-svn: 154191
the template instantiation of statement-expressions.
I think it was jyasskin who had a crashing testcase in this area;
hopefully this fixes it and he can find his testcase and check it in.
llvm-svn: 154189
The warning this inhibits, -Wobjc-root-class, is opt-in for now. However, all clang unit tests that would trigger
the warning have been updated to use -Wno-objc-root-class. <rdar://problem/7446698>
llvm-svn: 154187
inside of a sysroot targeting a system+sysroot which is "similar" or
"compatible" with the host system. This shows up when trying to build
system images on largely compatible hardware as-if fully cross compiled.
The problem is that previously we *perfectly* mimiced GCC here, and it
turns out GCC has a bug that no one has really stumbled across. GCC will
try to look in thy system prefix ('/usr/local' f.ex.) into which it is
instaled to find libraries installed along side GCC that should be
preferred to the base system libraries ('/usr' f.ex.). This seems not
unreasonable, but it has a very unfortunate consequence when combined
with a '--sysroot' which does *not* contain the GCC installation we're
using to complete the toolchain. That results in some of the host
system's library directories being searched during the link.
Now, it so happens that most folks doing stuff like this use
'--with-sysroot' and '--disable-multilib' when configuring GCC. Even
better, they're usually not cross-compiling to a target that is similar
to the host. As a result, searching the host for libraries doesn't
really matter -- most of the time weird directories get appended that
don't exist (no arm triple lib directory, etc). Even if you're
cross-compiling from 32-bit to 64-bit x86 or vice-versa, disabling
multilib makes it less likely that you'll actually find viable libraries
on the host. But that's just luck. We shouldn't rely on this, and this
patch disables looking in the system prefix containing the GCC
installation if that system prefix is *outside* of the sysroot. For
empty sysroots, this has no effect. Similarly, when using the GCC
*inside* of the sysroot, we still track wherever it is installed within
the sysroot and look there for libraries. But now we can use a cross
compiler GCC installation outside the system root, and only look for the
crtbegin.o in the GCC installation, and look for all the other libraries
inside the system root.
This should fix PR12478, allowing Clang to be used when building
a ChromiumOS image without polluting the image with libraries from the
host system.
llvm-svn: 154176
Richard Smith