Chris Lattner [Fri, 31 Aug 2012 22:39:21 +0000 (22:39 +0000)]
don't warn about unused values when the unused value is a statement expression expanded from a macro. This is of dubious utility in general, but is specifically a major issue for the linux kernel. This resolves PR13747.
Thread-safety analysis: fix handling of LOCK_RETURNED attribute so that the
latest definition of a function is always used when computing lock expressions.
Joao Matos [Fri, 31 Aug 2012 21:10:54 +0000 (21:10 +0000)]
Emulate MSVC's preprocessor macro argument separator behavior by not considering commas from nested macro expansions as argument separators. Fixes parsing of VS 2012 headers.
Joao Matos [Fri, 31 Aug 2012 18:45:21 +0000 (18:45 +0000)]
Improved MSVC __interface support by adding first class support for it, instead of aliasing to "struct" which had some incorrect behaviour. Patch by David Robins.
Jordan Rose [Fri, 31 Aug 2012 18:19:18 +0000 (18:19 +0000)]
[analyzer] RetainCountChecker: don't assume all functions have names.
Fixes a hard-to-reach crash when calling a non-member overloaded operator
with arguments that may be callbacks.
Future-proofing: don't make the same assumption in MallocSizeofChecker.
Aside from possibly respecting attributes in the future, it might be
possible to call 'malloc' through a function pointer.
I audited all other uses of FunctionDecl::getIdentifier() in the analyzer;
they all now correctly test to see if the identifier is present before
using it.
Joao Matos [Fri, 31 Aug 2012 17:28:09 +0000 (17:28 +0000)]
Fixed invalid iterators. Since the iterator is invalidated after the erase() call, we need to replace it with the new one returned. This was triggering an assert under MSVC 2012 and making all the ARCMT/ tests fail.
Jordan Rose [Fri, 31 Aug 2012 17:06:49 +0000 (17:06 +0000)]
[analyzer] Though C++ inlining is enabled, don't inline ctors and dtors.
More generally, this adds a new configuration option 'c++-inlining', which
controls which C++ member functions can be considered for inlining. This
uses the new -analyzer-config table, so the cc1 arguments will look like this:
Note that each mode implies that all the previous member function kinds
will be inlined as well; it doesn't make sense to inline destructors
without inlining constructors, for example.
Jordan Rose [Fri, 31 Aug 2012 00:36:26 +0000 (00:36 +0000)]
[analyzer] Ensure that PathDiagnostics profile the same regardless of path.
PathDiagnostics are actually profiled and uniqued independently of the
path on which the bug occurred. This is used to merge diagnostics that
refer to the same issue along different paths, as well as by the plist
diagnostics to reference files created by the HTML diagnostics.
However, there are two problems with the current implementation:
1) The bug description is included in the profile, but some
PathDiagnosticConsumers prefer abbreviated descriptions and some
prefer verbose descriptions. Fixed by including both descriptions in
the PathDiagnostic objects and always using the verbose one in the profile.
2) The "minimal" path generation scheme provides extra information about
which events came from macros that the "extensive" scheme does not.
This resulted not only in different locations for the plist and HTML
diagnostics, but also in diagnostics being uniqued in the plist output
but not in the HTML output. Fixed by storing the "end path" location
explicitly in the PathDiagnostic object, rather than trying to find the
last piece of the path when the diagnostic is requested.
This should hopefully finish unsticking our internal buildbot.
Eli Friedman [Fri, 31 Aug 2012 00:14:07 +0000 (00:14 +0000)]
Change the representation of builtin functions in the AST
(__builtin_* etc.) so that it isn't possible to take their address.
Specifically, introduce a new type to represent a reference to a builtin
function, and a new cast kind to convert it to a function pointer in the
operand of a call. Fixes PR13195.
objective-C ARC: under -Wexplicit-ownership-type diagnose those
method parameter types which are reference to an objective-C
pointer to object with no explicit ownership. // rdar://10907090
Douglas Gregor [Thu, 30 Aug 2012 21:47:37 +0000 (21:47 +0000)]
The presence of a user-*declared* constructor makes the default
constructor not user provided (and, therefore, non-trivial). Fixes
<rdar://problem/11736429>.
Jordan Rose [Thu, 30 Aug 2012 20:43:09 +0000 (20:43 +0000)]
[analyzer] Plist diagnostics: Fix a case where we fail to close an XML tag.
If the current path diagnostic does /not/ have files associated with it, we
were simply skipping on to the next diagnostic with 'continue'. But that
also skipped the close tag for the diagnostic's <dict> node.
Richard Trieu [Thu, 30 Aug 2012 20:32:24 +0000 (20:32 +0000)]
Add -Wduplicate-enum warning. Clang will emit this warning when an implicitly
initiated enum constant has the same value as another enum constant.
For instance:
enum test { A, B, C = -1, D, E = 1 };
Clang will warn that:
A and D both have value 0
B and E both have value 1
A few exceptions are made to keep the noise down. Enum constants which are
initialized to another enum constant, or an enum constant plus or minus 1 will
not trigger this warning. Also, anonymous enums are not checked.
Douglas Gregor [Thu, 30 Aug 2012 20:04:43 +0000 (20:04 +0000)]
Extend the "__is_pod" hack, which demotes various type trait keywords
(__is_pod, __is_signed, etc.) to normal identifiers if they are
encountered in certain places in the grammar where we know that prior
versions of libstdc++ or libc++ use them, to still allow the use of
these keywords as type traits. Fixes <rdar://problem/9836262> and PR10184.
Manuel Klimek [Thu, 30 Aug 2012 19:41:06 +0000 (19:41 +0000)]
Fixes a bug for binding memoized match results.
Intorduces an abstraction for DynTypedNode which makes
is impossible to create in ways that introduced the bug;
also hides the implementation details of the template
magic away from the user and prepares the code for adding
QualType and TypeLoc bindings, as well as using DynTypedNode
instead of overloads for child and ancestor matching.
getNodeAs<T> was changed towards a non-pointer type, as
we'll want QualType and TypeLoc nodes to be returned
by value (the alternative would be to create new storage
which is prohibitively costly if we want to use it for
child / ancestor matching).
DynTypedNode is moved into a new header ASTTypeTraits.h,
as it is completely independent of the rest of the matcher
infrastructure - if the need comes up, we can move it to
a more common place.
The interface for users before the introduction of the
common storage change remains the same, minus the introduced
bug, for which a regression test was added.
objective-C: clang must implicitly convert
__objc_yes/__objc_no to (BOOL)1/(BOOL)0 when
BOOL is declared; otherwise it resorts to
default of 'signed char'. This is important to
selecting the correct Numeric API numberWithBool:
Can't have a clang test for this. Will checkin and
executable llvm test. // rdar://12156616
Richard Smith [Thu, 30 Aug 2012 13:38:46 +0000 (13:38 +0000)]
Make preprocessor act in a GCC-compatible fashion when a macro is redefined
within its own argument list. The original definition is used for the immediate
expansion, but the new definition is used for any subsequent occurences within
the argument list or after the expansion.
Richard Smith [Thu, 30 Aug 2012 13:13:20 +0000 (13:13 +0000)]
PR13652: Don't assume the parameter array on a FunctionTypeLoc for a lambda will
be filled in; they won't if the lambda's declarator has an invalid type. Instead
take the parameters from the declarator directly.
Bill Wendling [Thu, 30 Aug 2012 00:43:41 +0000 (00:43 +0000)]
Emit .gcda files as absolute paths instead of relative paths.
This improves compatibility with gcc in this regard, and this file generation
can be ameliorated with GCOV_PREFIX and GCOV_PREFIX_STRIP. It's also useful if
your build directory doesn't specify -o <abspath> and it uses a recursive make
structure, so it's not relative to the toplevel.
Patch by Joshua Cranmer!
<rdar://problem/12179524>
Anna Zaks [Wed, 29 Aug 2012 23:23:43 +0000 (23:23 +0000)]
[analyzer] Stop tracking symbols based on a retain count summary of
inlined function.
This resolves retain count checker false positives that are caused by
inlining ObjC and other methods. Essentially, if we are passing an
object to a method with "delegate" in the selector or a function pointer
as another argument, we should stop tracking the other parameters/return
value as far as the retain count checker is concerned.
Anna Zaks [Wed, 29 Aug 2012 21:22:37 +0000 (21:22 +0000)]
[analyzer] Improved diagnostic pruning for calls initializing values.
This heuristic addresses the case when a pointer (or ref) is passed
to a function, which initializes the variable (or sets it to something
other than '0'). On the branch where the inlined function does not
set the value, we report use of undefined value (or NULL pointer
dereference). The access happens in the caller and the path
through the callee would get pruned away with regular path pruning. To
solve this issue, we previously disabled diagnostic pruning completely
on undefined and null pointer dereference checks, which entailed very
verbose diagnostics in most cases. Furthermore, not all of the
undef value checks had the diagnostic pruning disabled.
This patch implements the following heuristic: if we pass a pointer (or
ref) to the region (on which the error is reported) into a function and
it's value is either undef or 'NULL' (and is a pointer), do not prune
the function.
Simon Atanasyan [Wed, 29 Aug 2012 19:59:32 +0000 (19:59 +0000)]
Use getTargetDefines() virtual function in MipsTargetInfoBase successors
to define all macros for MIPS targets. Remove redundant virtual function
getArchDefines(). Two virtual functions for this task are really too much.
Hans Wennborg [Wed, 29 Aug 2012 18:27:29 +0000 (18:27 +0000)]
Move TLS check from LValueExprEvaluator::VisitVarDecl to
CheckLValueConstantExpression.
Richard pointed out that using the address of a TLS variable is ok in a
core C++11 constant expression, as long as it isn't part of the eventual
result of constant expression evaluation. Having the check in
CheckLValueConstantExpression accomplishes this.
Fixed a problem with #pragma push_macro/pop_macro implementation.
Summary:
The problem was with the following sequence:
#pragma push_macro("long")
#undef long
#pragma pop_macro("long")
in case when "long" didn't represent a macro.
Fixed crash and removed code duplication for #undef/pop_macro case. Added regression tests.
Ted Kremenek [Wed, 29 Aug 2012 05:55:00 +0000 (05:55 +0000)]
Add new -cc1 driver option -analyzer-config, which allows one to specify
a comma separated collection of key:value pairs (which are strings). This
allows a general way to provide analyzer configuration data from the command line.
Jordan Rose [Wed, 29 Aug 2012 01:11:59 +0000 (01:11 +0000)]
[analyzer] C++ objects returned on the stack may be wrapped in ExprWithCleanups.
In C++, objects being returned on the stack are actually copy-constructed into
the return value. That means that when a temporary is returned, it still has
to be destroyed, i.e. the returned expression will be wrapped in an
ExprWithCleanups node. Our "returning stack memory" checker needs to look
through this node to see if we really are returning an object by value.
Summary:
Summary: Keep history of macro definitions and #undefs with corresponding source locations, so that we can later find out all macros active in a specified source location. We don't save the history in PCH (no need currently). Memory overhead is about sizeof(void*)*3*<number of macro definitions and #undefs>+<in-memory size of all #undef'd macros>
I've run a test on a file composed of 109 .h files from boost 1.49 on x86-64 linux.
Stats before this patch:
*** Preprocessor Stats:
73222 directives found:
19171 #define.
4345 #undef.
#include/#include_next/#import:
5233 source files entered.
27 max include stack depth
19210 #if/#ifndef/#ifdef.
2384 #else/#elif.
6891 #endif.
408 #pragma.
14466 #if/#ifndef#ifdef regions skipped
80023/451669/1270 obj/fn/builtin macros expanded, 85724 on the fast path.
127145 token paste (##) operations performed, 11008 on the fast path.
Stats with this patch:
...
Preprocessor Memory: 7541687B total
BumpPtr: 6066176
Macro Expanded Tokens: 417768
Predefines Buffer: 8135
Macros: 1048576
#pragma push_macro Info: 0
Poison Reasons: 1024
Comment Handlers: 8
In my test increase in memory usage is about 1.7Mb, which is ~28% of initial preprocessor's memory usage and about 0.8% of clang's total VMM allocation.
As for CPU overhead, it should only be noticeable when iterating over all macros, and should mostly consist of couple extra dereferences and one comparison per macro + skipping of #undef'd macros. It's less trivial to measure, though, as the preprocessor consumes a very small fraction of compilation time.
Manuel Klimek [Tue, 28 Aug 2012 23:26:39 +0000 (23:26 +0000)]
Modifes BoundNodes to store void* and allow casting them
into the correct types when pulling them out in the result
callback in a type safe way.
This is also the base change for multiple things that will
allow handling things more generally and thus supporting more
of the AST, especially handling Type nodes.
objective-C arc: ns_returns_retained is a type attribute in ARC,
and when used in property type declaration, is handled as type
attribute. Do not issue the warning when declaraing the property.
// rdar://12173491
Aaron Ballman [Tue, 28 Aug 2012 20:55:40 +0000 (20:55 +0000)]
Splitting the duplicated decl spec extension warning into two: one is an ExtWarn and the other a vanilla warning. This addresses PR13705, where const char const * wouldn't warn unless -pedantic was specified under the right conditions.
Jordan Rose [Tue, 28 Aug 2012 20:52:21 +0000 (20:52 +0000)]
[analyzer] Teach CallEventManager that CXXTemporaryObjectExpr is also a ctor.
Specifically, CallEventManager::getCaller was looking at the call site for
an inlined call and trying to see what kind of call it was, but it only
checked for CXXConstructExprClass. (It's not using an isa<> here to avoid
doing three more checks on the the statement class.)
This caused an unreachable when we actually did inline the constructor of a
temporary object.
Jordan Rose [Tue, 28 Aug 2012 20:52:13 +0000 (20:52 +0000)]
[analyzer] When we look for the last stmt in a function, skip implicit dtors.
When exiting a function, the analyzer looks for the last statement in the
function to see if it's a return statement (and thus bind the return value).
However, the search for "the last statement" was accepting statements that
were in implicitly-generated inlined functions (i.e. destructors). So we'd
go and get the statement from the destructor, and then say "oh look, this
function had no explicit return...guess there's no return value". And /that/
led to the value being returned being declared dead, and all our leak
checkers complaining.
CUDA: give static storage class to __shared__ and __constant__
variables without a storage class within a function, to implement
CUDA B.2.5: "__shared__ and __constant__ variables have implied static
storage [duration]."
projects / clang / log