Simon Atanasyan [Mon, 27 Aug 2012 20:55:56 +0000 (20:55 +0000)]
MIPS: Use -G option to specify MIPS section threshold. Translate it
to the -mllvm -mips-ssection-threshold=<value> pair and pass to the frontend.
The patch suggested by Carl Norum.
Chad Rosier [Mon, 27 Aug 2012 20:23:31 +0000 (20:23 +0000)]
[ms-inline asm] Rename GenerateAsmString to generateAsmString to conform with
coding standards. Also, add stub for MSAsmStmt class as part of unifying
codegen logic for AsmStmts.
Manuel Klimek [Mon, 27 Aug 2012 18:49:12 +0000 (18:49 +0000)]
This is a temporary solution until we have a better way to
parse doxygen comments for macros with libclang.
I'm not entirely happy about this script, but as it saves
a lot of work in keeping the docs up to date with the
actual code I think checking it in makes sense.
Jordan Rose [Mon, 27 Aug 2012 18:39:22 +0000 (18:39 +0000)]
[analyzer] Don't inline constructors for objects allocated with operator new.
Because the CXXNewExpr appears after the CXXConstructExpr in the CFG, we don't
actually have the correct region to construct into at the time we decide
whether or not to inline. The long-term fix (discussed in PR12014) might be to
introduce a new CFG node (CFGAllocator) that appears before the constructor.
Tracking the short-term fix in <rdar://problem/12180598>.
Jordan Rose [Mon, 27 Aug 2012 17:50:07 +0000 (17:50 +0000)]
[analyzer] Inline constructors for any object with a trivial destructor.
This allows us to better reason about status objects, like Clang's own
llvm::Optional (when its contents are trivially destructible), which are
often intended to be passed around by value.
We still don't inline constructors for temporaries in the general case.
Jordan Rose [Sat, 25 Aug 2012 01:06:23 +0000 (01:06 +0000)]
[analyzer] Use the common evalBind infrastructure for initializers.
This allows checkers (like the MallocChecker) to process the effects of the
bind. Previously, using a memory-allocating function (like strdup()) in an
initializer would result in a leak warning.
This does bend the expectations of checkBind a bit; since there is no
assignment expression, the statement being used is the initializer value.
In most cases this shouldn't matter because we'll use a PostInitializer
program point (rather than PostStmt) for any checker-generated nodes, though
we /will/ generate a PostStore node referencing the internal statement.
(In theory this could have funny effects if someone actually does an
assignment within an initializer; in practice, that seems like it would be
very rare.)
Richard Smith [Sat, 25 Aug 2012 00:32:28 +0000 (00:32 +0000)]
-fcatch-undefined-behavior: add the -ftrapv checks to the set of things caught
by this mode, and also check for signed left shift overflow. The rules for the
latter are a little subtle:
* neither C89 nor C++98 specify the behavior of a signed left shift at all
* in C99 and C11, shifting a 1 bit into the sign bit has undefined behavior
* in C++11, with core issue 1457, shifting a 1 bit *out* of the sign bit has
undefined behavior
As of this change, we use the C99 rules for all C language variants, and the
C++11 rules for all C++ language variants. Once we have individual
-fcatch-undefined-behavior= flags, this should be revisited.
objective-C: Do not warn if align attribute on method
declaration is not provided. It is only necessary on
the method implementation. // rdar://11593375
John McCall [Fri, 24 Aug 2012 22:54:02 +0000 (22:54 +0000)]
When computing the effective context for access control,
make sure we walk up the DC chain for the current context,
rather than allowing ourselves to get switched over to the
canonical DC chain. Fixes PR13642.
Chad Rosier [Fri, 24 Aug 2012 21:42:51 +0000 (21:42 +0000)]
[ms-inline asm] Change the -fenable-experimental-ms-inline-asm option from a
CodeGen option to a LangOpt option. In turn, hoist the guard into the parser
so that we avoid the new (and fairly unstable) Sema/AST/CodeGen logic. This
should restore the behavior of clang to that prior to r158325.
<rdar://problem/12163681>
objective-C: When checking for valid overriden property
in class extension, assume default is rewdwrite and don't
issue any diagnostics, privided other ownership models
are ok.
Ted Kremenek [Fri, 24 Aug 2012 19:46:03 +0000 (19:46 +0000)]
Rename the "experimental" checker package to "alpha". We will then refine
this group into "alpha" and "beta" to distinguish between checkers in
different levels of premature state.
Ted Kremenek [Fri, 24 Aug 2012 19:35:19 +0000 (19:35 +0000)]
Rework how PathDiagnosticConsumers pass knowledge of what files they
generated for a given diagnostic to another. Because PathDiagnostics
are specific to a give PathDiagnosticConsumer, store in
a FoldingSet a unique hash for a PathDiagnostic (that will be the same
for the same bug for different PathDiagnosticConsumers) that
stores a list of files generated. This can then be read by the
other PathDiagnosticConsumers.
Jordan Rose [Fri, 24 Aug 2012 16:34:31 +0000 (16:34 +0000)]
[analyzer] If we dereference a NULL that came from a function, show the return.
More generally, any time we try to track where a null value came from, we
should show if it came from a function. This usually isn't necessary if
the value is symbolic, but if the value is just a constant we previously
just ignored its origin entirely. Now, we'll step into the function and
recursively add a visitor to the returned expression.
James Dennett [Fri, 24 Aug 2012 06:59:51 +0000 (06:59 +0000)]
Allow RecursiveASTVisitor to visit CXXCtorInitializer objects for which
isWritten() returns false, if shouldVisitImplicitCode() returns true.
Previously those CXXCtorInitializers were always skipped.
In order to make this change easier to test, this patch also extends the
test class template ExpectedLocationVisitor to support arbitrary numbers
of expected matches and disallowed matches.
Ted Kremenek [Fri, 24 Aug 2012 06:49:34 +0000 (06:49 +0000)]
Go ahead and show experimental checkers in the scan-build "-h" output.
They are labeled as not being enabled-by-default, and how else
are users going to test them.
Daniel Jasper [Fri, 24 Aug 2012 05:12:34 +0000 (05:12 +0000)]
Rename the ASTMatchers to better match AST nodes. Now, all
ASTMatchers have the same name as the corresponding AST nodes
but are lower case. The only exceptions are the "CXX" prefixes
which are not copied over to the matcher names as the goal is to
actually remove these prefixes from the AST node names.
Anna Zaks [Fri, 24 Aug 2012 01:39:13 +0000 (01:39 +0000)]
[analyzer] Remove unnecessary code.
This code has been added a while ago and removing it does not trigger
any test failures. The false positives it was trying to suppress are
probably handled by other logic (ex: special handling of delegates).
Richard Smith [Fri, 24 Aug 2012 00:54:33 +0000 (00:54 +0000)]
New -fcatch-undefined-behavior features:
* when checking that a pointer or reference refers to appropriate storage for a type, also check the alignment and perform a null check
* check that references are bound to appropriate storage
* check that 'this' has appropriate storage in member accesses and member function calls
Anna Zaks [Fri, 24 Aug 2012 00:06:12 +0000 (00:06 +0000)]
[analyzer] Make analyzer less aggressive when dealing with [self init].
With inlining, retain count checker starts tracking 'self' through the
init methods. The analyser results were too noisy if the developer
did not follow 'self = [super init]' pattern (which is common
especially in older code bases) - we reported self init anti-pattern AND
possible use-after-free. This patch teaches the retain count
checker to assume that [super init] does not fail when it's not consumed
by another expression. This silences the retain count warning that warns
about possibility of use-after-free when init fails, while preserving
all the other checking on 'self'.
Jordan Rose [Thu, 23 Aug 2012 23:16:34 +0000 (23:16 +0000)]
[scan-build] Accept -fno-objc-arc as well as -fobjc-arc.
This is how Xcode lets individual files be marked as non-ARC when the rest
of the project is ARC-enabled, so this is necessary for scan-build xcodebuild.
Jordan Rose [Thu, 23 Aug 2012 23:01:43 +0000 (23:01 +0000)]
[analyzer] For now, treat pointers-to-members as non-null void * symbols.
Until we have full support for pointers-to-members, we can at least
approximate some of their use by tracking null and non-null values.
We thus treat &A::m_ptr as a non-null void * symbol, and MemberPointer(0)
as a pointer-sized null constant.
This enables support for what is sometimes called the "safe bool" idiom,
demonstrated in the test case.
Jordan Rose [Thu, 23 Aug 2012 23:01:39 +0000 (23:01 +0000)]
[analyzer] Handle UserDefinedConversion casts in C++.
This is trivial; the UserDefinedConversion always wraps a CXXMemberCallExpr
for the appropriate conversion function, so it's just a matter of
propagating that value to the CastExpr itself.
Dmitri Gribenko [Thu, 23 Aug 2012 22:40:40 +0000 (22:40 +0000)]
Attaching comments to decls: since it was decided that Decl::isImplicit should
not be set for implicit instantiations, remove the FIXME. This should be the
last bit for PR13634. The actual fix happened in r162238.
Motivation: it might be misleading to mark implicit instantiations as
Decl::isImplicit = true. Because then, in order to be consistent, we should
mark all instantiated members as implicit. But the user did actually type the
declaration for the member, but the compiler played with it a little bit.
Chad Rosier [Thu, 23 Aug 2012 21:55:11 +0000 (21:55 +0000)]
[ms-inline asm] Add a few helper function to the MSAsmStmt class that are needed
by CodeGen.
In the long-term, much of the codegen logic will be shared between the GNU-style
and MS-style inline assembly, but for now I'm replicating this logic to avoid
regressions with the GNU-style.
Ted Kremenek [Thu, 23 Aug 2012 20:46:57 +0000 (20:46 +0000)]
Change a bunch of cases where we do "getAs<...>->doSomething()" to
"castAs<...>->doSomething()". The analyzer was flagging these
as potential null dereferences, which is technically true. The
invariants appear to be that these casts should never fail, so
let's use castAs<> instead and avoid a runtime check.
Richard Smith [Thu, 23 Aug 2012 20:19:14 +0000 (20:19 +0000)]
When disambiguating an expression-statement from a declaraton-statement, if the
statement starts with an identifier for which name lookup will fail either way,
look at later tokens to disambiguate in order to improve error recovery.
Jordan Rose [Thu, 23 Aug 2012 18:10:53 +0000 (18:10 +0000)]
[analyzer] Support C++ default arguments if they are literal values.
A CXXDefaultArgExpr wraps an Expr owned by a ParmVarDecl belonging to the
called function. In general, ExprEngine and Environment ought to treat this
like a ParenExpr or other transparent wrapper expression, with the inside
expression evaluated first.
However, if we call the same function twice, we'd produce a CFG that contains
the same wrapped expression twice, and we're not set up to handle that. I've
added a FIXME to the CFG builder to come back to that, but meanwhile we can
at least handle expressions that don't need to be explicitly evaluated:
literals. This probably handles many common uses of default parameters:
true/false, null, etc.
Douglas Gregor [Thu, 23 Aug 2012 05:05:18 +0000 (05:05 +0000)]
array_pod_sort on the addresses of declaration pointers leads to
inconsistent ordering of results; instead, use use SmallPtrSet to
eliminate duplicates.
Richard Smith [Thu, 23 Aug 2012 04:53:18 +0000 (04:53 +0000)]
Fix undefined behavior: don't call ObjCInterfaceDecl::isArcWeakrefUnavailable
on a null pointer. (This function happens to work for a null 'this' pointer, so
no test.)
Eli Friedman [Thu, 23 Aug 2012 03:10:17 +0000 (03:10 +0000)]
Use the alignment from lvalue emission to more accurately compute the alignment
of a pointer for builtin emission, instead of just depending on the type of the
pointee. <rdar://problem/11314941>.
projects / clang / log