need to provide a 'dominating IP' which is guaranteed to
dominate the (de)activation point but which cannot be avoided
along any execution path from the (de)activation point to
the push-point of the cleanup. Using the entry block is
bad mojo.
llvm-svn: 144276
a previously-inactive cleanup, not only do we need a
flag variable, but we should also force the cleanup to
query the flag variable. However, we only need to do
this when we're activating in a context that's
conditionally executed; otherwise, we may safely
assume that the cleanup is dominated by the activation
point.
llvm-svn: 144271
full-expression. Naturally they're inactive before we enter
the block literal expression. This restores the intended
behavior that blocks belong to their enclosing scope.
There's a useful -O0 / compile-time optimization that we're
missing here with activating cleanups following straight-line
code from their inactive beginnings.
llvm-svn: 144268
version of Ubuntu. It has a very broken multiarch configuration, and so
we need special logic to handle it correctly. Fixing and testing this
uncovered a few other trivial issues with the logic that are fixed as
well.
I added tests to cover this as it is hard to notice if you install
recent versions of the OS.
llvm-svn: 144165
block-typed __block variables using objc_retainBlock
and objc_dispose. Previously we were using
_Block_object_assign and _Block_object_destroy
with BLOCK_BYREF_CALLER, which causes the runtime
to completely ignore the retain and release.
In most cases this doesn't cause catastrophe
because the retain/release are balanced and
because the block in the variable was copied
upon assignment there. However, the stack
copy of the variable will be released when
it goes out of scope, which is a problem if
that value was released due to an assignment
to the heap copy. Similarly, a leak can occur
if the variable is assigned after the copy to
the heap.
llvm-svn: 144162
In certain cases ASTReader would call the normal DiagnosticsEngine API to initialize
the state of diagnostic pragmas but DiagnosticsEngine would try to compare source locations
leading to crash because the main FileID was not yet initialized.
Yet another case of the ASTReader trying to use the normal APIs and inadvertently breaking
invariants. Fix this by having the ASTReader set up the internal state directly.
llvm-svn: 144153
which they do. This avoids all of the default argument promotions that
we (1) don't want, and (2) undo during that custom type checking, and
makes sure that we don't run into trouble during template
instantiation. Fixes PR11320.
llvm-svn: 144110
The Neon load/store intrinsics need to be implemented as macros to avoid
hiding alignment attributes on the pointer arguments, and the macros can
only evaluate those pointer arguments once (in case they have side effects),
so it has been hard to get the right type checking for those pointers.
I tried various alternatives in the arm_neon.h header, but it's much more
straightforward to just check directly in Sema.
llvm-svn: 144075
expression evaluation:
- When folding a non-value-dependent expression, we may try to use the
initializer of a value-dependent variable. If that happens, give up.
- In C++98, actually check that a const, non-volatile DeclRefExpr inside an ICE
is of integral or enumeration type (a reference isn't OK!)
- In C++11, DeclRefExprs for objects of const literal type initialized with
value-dependent expressions are themselves value-dependent.
- So are references initialized with value-dependent expressions (though this
case is missing from the C++11 standard, along with many others).
llvm-svn: 144056
doesn't duplicate, but they all surface as implicit
properties. It's also a useful optimization to not
duplicate the implicit getter lookup. So, trust the
getter lookup that was already done in these cases.
llvm-svn: 144031
initializer; all other constexpr variables are merely required to be
initialized. In particular, a user-provided constexpr default constructor can be
used for such initialization.
llvm-svn: 144028
default", make a note of which is used when creating the
initial declaration. Previously, we would wait until later to handle
default/delete as a definition, but this is too late: when adding the
declaration, we already treated the declaration as "user-provided"
when in fact it was merely "user-declared".
Fixes PR10861 and PR10442, along with a bunch of FIXMEs.
llvm-svn: 144011
-Move __strong/__weak added to a property type to the property attribute,
e.g. "@property (assign) __weak Foo *prop;" --> "@property (weak) Foo *prop;"
-Remove (assign) in a property so that it becomes strong-by-default in ARC.
llvm-svn: 143979
__int128_t and __uint128_t. Short and unsigned short integer literals support
is only to work around a crasher as reported in PR11179 and will be removed
once Clang no longer builds short integer literals.
llvm-svn: 143977
__weak is unsupported by the deployment target, since it is going to be
ignored anyway.
Makes it easier for incremental migration from GC.
llvm-svn: 143975
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).
Fixes <rdar://problem/9676205>.
Reinstated, now that we have the fix in r143967.
llvm-svn: 143968
We don't actually need a separate flag for non-sysrooted paths as the
driver has to manage the sysroot anyways. The driver is not infrequently
adding paths to the header search based on their existence on the
filesystem. For that, it has to add the sysroot anyways, we should pass
it on down to CC1 already joined. More importantly, the driver cannot in
all cases distinguish between sysrooted paths and paths that are
relative to the Clang binary's installation directory. Essentially, we
always need to ignore the system root for these internal header search
options. It turns out in most of the places we were already providing
the system root in the driver, and then another one in CC1 so this fixes
several bugs.
llvm-svn: 143917
partially undoes the revert in r143491, but does not introduce any new instances
of the underlying issue (which is not yet fixed) in code which does not use
the 'constexpr' keyword.
llvm-svn: 143905
property references to use a new PseudoObjectExpr
expression which pairs a syntactic form of the expression
with a set of semantic expressions implementing it.
This should significantly reduce the complexity required
elsewhere in the compiler to deal with these kinds of
expressions (e.g. IR generation's special l-value kind,
the static analyzer's Message abstraction), at the lower
cost of specifically dealing with the odd AST structure
of these expressions. It should also greatly simplify
efforts to implement similar language features in the
future, most notably Managed C++'s properties and indexed
properties.
Most of the effort here is in dealing with the various
clients of the AST. I've gone ahead and simplified the
ObjC rewriter's use of properties; other clients, like
IR-gen and the static analyzer, have all the old
complexity *and* all the new complexity, at least
temporarily. Many thanks to Ted for writing and advising
on the necessary changes to the static analyzer.
I've xfailed a small diagnostics regression in the static
analyzer at Ted's request.
llvm-svn: 143867
path. That assumption should never have been true, but it was until
I fixed it. Now that its fixed, add a triple here to get correct
behavior even on Windows.
llvm-svn: 143863
edge cases and have better behavior. Specifically, we should actually
prefer the general '4.6' version string over the '4.6.1' string, as
'4.6.2' should be able to replace it without breaking rpaths or any
other place that these paths have been embedded. Debian-based
distributions are already using a path structure with symlinks to
achieve in-place upgrades for patch versions. Now our parsing reflects
this and we select the shorter paths instead of the longer paths.
A separate issue was that we would not parse a leading patch version
number even in the presence of a suffix. The above change makes this
more problematic as it would cause a suffix being added to make us treat
the entire thing as patch-version-agnostic, which it isn't. This changes
the logic to distinguish between '4.4.x' and 4.4.1-x', and retain that
the latter has *some* patch number information. Currently, we always
bias toward the shorter and more canonical version strings. If it
becomes important we can add more Debian like rules to produce sequences
such as '4.4.1b' > '4.4.1' > '4.4.1-rc3' > '4.4.1-rc2' > '4.4.1-pre5',
but I'm very doubtful this will ever matter or be desirable.
I've made the tests for this logic a bit more interesting, and added
some specific tests for logic that is now different.
llvm-svn: 143841
rather than presuming that it is 3.0. This is extra important as the
version should be 3.1, but CMake hasn't caught up with the times.
That'll be fixed in a separate commit.
llvm-svn: 143823
builtin headers are no longer going to receive the old 'implicit extern
"C" block' semantics. This hint is actually ignored by both Clang and
GCC at this point, and Clang's own builtin headers can simply be changed
if there is any issue with this. Clang should be free to include these
however it wants, and so shorter and simpler is better.
Note: *nothing* is changing about the *system* stddef.h include. That
should always have the exact same include semantics, whether with Clang
or GCC or any other compiler. Only the compiler-builtin header search
path is changing.
If anyone knows of some risk that this introduces that I've not thought
of, please chime in. So far, only Windows has switched to the Brave New
World, but others should be switching soon.
llvm-svn: 143806
to do "realistic" includes, and so need the header search logic now in
the driver. This in turn requires switching the CC1 options to the
actual driver options, and passing -Xclang where there is no analogy.
llvm-svn: 143805
is a pretty gross hack, but I don't have any significantly cleaner ideas
for this. There are several things obviously gross about it:
1) Lit shouldn't know that Clang needs this. This really that bad, as
Lit already knows about CC1 and other internal details.
2) This hard codes the '3.0' version number, which is pretty lame.
3) This hard codes every other aspect of the resource dir structure
which is less lame than the version number, but still not great.
However, it should bring the MSVC tests back to life, and it should
unblock the rest of the move from Frontend to Driver, so I think it's
worth a bit of grossness that is isolated in our testing infrastructure
while we figure out the best long term approach. I have the following
ideas, some of which only solve part of the problem (and thus might need
to be combined with other ideas):
a) Create a symlink or other convenience path instead of a version
number.
b) Run 'clang' directly in the lit.cfg, look at its resource dir, and use
that.
c) Switch all the tests to use the driver instead of CC1.
d) Hack the frontend to synthesize builtin include directories when none
are provided by the driver.
I don't like (d) because it feels very hackish and likely to break. We
can only solve a small part of the problem with (a). I wanted to vote
for (c), but lots of the tests in this bucket are really heavily using
internal-only flags like -verify and -triple. I'm loath to complicate
them with the full driver layer. Also, switching them to the driver adds
more than just builtin headers, but all of the rest of the system
headers!
This leaves me with (b). If others like (b), I'll switch to it, but it
felt a bit icky. Nothing concrete, and the other options look
significantly worse, but I felt icky enough that I wanted to start with
a more brain-dead patch to stop the bleeding, and gauge others' feelings
here.
llvm-svn: 143804
to allow us to implement the C++11 rule that a non-active union member can't be
read, and use it to implement subobject access for string literals.
llvm-svn: 143677
A PCH file keeps track of #pragma diagnostics state; when loading the preamble, they conflicted
with the #pragma diagnostic state already present in the DiagnosticsEngine object due to
parsing the preamble.
Fix this by clearing the state of the DiagnosticsEngine object.
Fixes rdar://10363572 && http://llvm.org/PR11254.
llvm-svn: 143644
definition, we may not have a scope corresponding to the namespace
where that friend function template actually lives. Work around this
issue by faking up a scope with the appropriate DeclContext.
This is a bit of a hack, but it fixes <rdar://problem/10204947>.
llvm-svn: 143614
to types. Enable this flag for code completion, where knowing whether
something is in an anonymous or inline namespace is actually not
useful, since you don't have to type it anyway. Fixes
<rdar://problem/10208818>.
llvm-svn: 143599
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).
Fixes <rdar://problem/9676205>.
llvm-svn: 143551
wrong class, make sure to drop it immediately; we don't want that
constructor to be available within the DeclContext. Fixes
<rdar://problem/9677163>.
llvm-svn: 143506
does not match any declaration in the class (or class template), be
sure to mark it as invalid. Fixes PR10924 / <rdar://problem/10119422>.
llvm-svn: 143504
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
llvm-svn: 143464
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).
llvm-svn: 143410
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.
llvm-svn: 143298
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.
llvm-svn: 143296
constexpr function arguments outside of their function (passing or returning
them by reference) does not work correctly yet.
Calling constexpr function templates does not work yet, since the bodies are not
instantiated until the end of the translation unit.
llvm-svn: 143234
are present in all the necessary places:
In constant expression evaluation, evaluate lvalues as lvalues and rvalues as
rvalues. Remove special case for caching reference initialization and fix a
cyclic initialization crash in the process.
llvm-svn: 143204
string is part of the function call, then there is no difference. If the
format string is not, the warning will point to the call site and a note
will point to where the format string is.
Fix-it hints for strings are moved to the note if a note is emitted. This will
prevent changes to format strings that may be used in multiple places.
llvm-svn: 143168
rvalue. An assertion to catch this is in ImpCastExprToType will follow, but
vector operations currently trip over this (due to omitting the usual arithmetic
conversions). Also add an assert to catch missing lvalue-to-rvalue conversions
on the LHS of ->.
llvm-svn: 143155
The OpenCL single precision division operation is only required to
be accurate to 2.5ulp. Annotate the fdiv instruction with metadata
which signals to the backend that an imprecise divide instruction
may be used.
llvm-svn: 143136
AST file more lazy, so that we don't eagerly load that information for
all known identifiers each time a new AST file is loaded. The eager
reloading made some sense in the context of precompiled headers, since
very few identifiers were defined before PCH load time. With modules,
however, a huge amount of code can get parsed before we see an
@import, so laziness becomes important here.
The approach taken to make this information lazy is fairly simple:
when we load a new AST file, we mark all of the existing identifiers
as being out-of-date. Whenever we want to access information that may
come from an AST (e.g., whether the identifier has a macro definition,
or what top-level declarations have that name), we check the
out-of-date bit and, if it's set, ask the AST reader to update the
IdentifierInfo from the AST files. The update is a merge, and we now
take care to merge declarations before/after imports with declarations
from multiple imports.
The results of this optimization are fairly dramatic. On a small
application that brings in 14 non-trivial modules, this takes modules
from being > 3x slower than a "perfect" PCH file down to 30% slower
for a full rebuild. A partial rebuild (where the PCH file or modules
can be re-used) is down to 7% slower. Making the PCH file just a
little imperfect (e.g., adding two smallish modules used by a bunch of
.m files that aren't in the PCH file) tips the scales in favor of the
modules approach, with 24% faster partial rebuilds.
This is just a first step; the lazy scheme could possibly be improved
by adding versioning, so we don't search into modules we already
searched. Moreover, we'll need similar lazy schemes for all of the
other lookup data structures, such as DeclContexts.
llvm-svn: 143100
The code had it backwards, thinking size_t was signed, and using that for "%zd".
Also let the analysis get the types for (u)intmax_t while we are at it.
llvm-svn: 143099
itself via an asm label.
available_externally functions are supposed to correspond to an external
function, and that is not the case in the examples in pr9614.
llvm-svn: 143049
essence, the redeclaration chain for a class could end up in an
inconsistent state while deserializing multiple declarations in that
chain, where the circular linked list was not, in fact,
circular. Since only two redeclarations of the same entity will get
loaded when we're in this state, restore circularity when both have
been loaded. Fixes <rdar://problem/10324940> / PR11195.
llvm-svn: 143037
expressions: expressions which refer to a logical rather
than a physical l-value, where the logical object is
actually accessed via custom getter/setter code.
A subsequent patch will generalize the AST for these
so that arbitrary "implementing" sub-expressions can
be provided.
Right now the only client is ObjC properties, but
this should be generalizable to similar language
features, e.g. Managed C++'s __property methods.
llvm-svn: 142914
Microsoft __if_exists/__if_not_exists statement. Also note that we
weren't traversing DeclarationNameInfo *at all* within the
RecursiveASTVisitor, which would be rather fatal for variadic
templates.
llvm-svn: 142906
statements. As noted in the documentation for the AST node, the
semantics of __if_exists/__if_not_exists are somewhat different from
the way Visual C++ implements them, because our parsed-template
representation can't accommodate VC++ semantics without serious
contortions. Hopefully this implementation is "good enough".
llvm-svn: 142901
analysis to separate dependent names from non-dependent names. For
dependent names, we'll behave differently from Visual C++:
- For __if_exists/__if_not_exists at class scope, we'll just warn
and then ignore them.
- For __if_exists/__if_not_exists in statements, we'll treat the
inner statement as a compound statement, which we only instantiate
in templates where the dependent name (after instantiation)
exists. This behavior is different from VC++, but it's as close as
we can get without encroaching ridiculousness.
The latter part (dependent statements) is not yet implemented.
llvm-svn: 142864
rvalues, as C++11 constant evaluation semantics require. DeclRefs referring to
references can now use the normal initialization-caching codepath, which
incidentally fixes a crash in cyclic initialization of references.
llvm-svn: 142844
Former RUN line does not make sense on Win32 hosts.
Win32 hosts would not be expected to set appropriate $PWD.
Latter RUN line might be made valid if PathV2::is_absolute("/foo") would be true. Unfortunately, "/foo" should not be treated as absolute path on Win32.
FYI, on mingw32 with MSYS bash (it has 'shell' feature);
Former) $PWD is set as "X:/hogehoge/test/Driver"
Latter) PWD=/foo sets "X:/root/to/msys/foo" to $PWD.
llvm-svn: 142721
in such a case just write out a reference of a previously serialized Stmt, instead
of serializing it all over again.
This saves memory + space + [de]serializing time, and avoids blowing up memory
with pathological cases. rdar://10293911
llvm-svn: 142696
class declaration which forces any such class and any
class that inherits from such a class to have their
typeinfo symbols be marked as weak.
// rdar://10246395
A test/CodeGenCXX/weak-extern-typeinfo.cpp
M lib/Sema/SemaDeclCXX.cpp
M lib/Sema/SemaDeclAttr.cpp
M lib/CodeGen/CGRTTI.cpp
llvm-svn: 142693
be sure to consider all of the possible lookup results. We were
assert()'ing (but behaving correctly) for unresolved values. Fixes
PR11134 / <rdar://problem/10290422>.
llvm-svn: 142652
but trivially constructible and destructible variables in C++11 mode. Also
incidentally improve the precision of the wording for jump diagnostics in C++98
mode.
llvm-svn: 142619
actually just has an extraneous 'template<>' header, strip off the
'template<>' header and treat it as a normal friend tag. Fixes PR10660
/ <rdar://problem/9958322>.
llvm-svn: 142587
instead of a semicolon (as sometimes happens during refactorings). When such a
comma is seen at the end of a line, and is followed by something which can't
possibly be a declarator (or even something which might be a plausible typo for
a declarator), suggest that a semicolon was intended.
llvm-svn: 142544
John McCall