John McCall [Tue, 8 May 2012 21:41:25 +0000 (21:41 +0000)]
Bind cleanups after doing l2r conversion on the operand of a
@throw expression; l2r conversion can introduce new cleanups
in certain cases, like when the expression is an ObjC property
reference of retainable type in ARC.
Anna Zaks [Tue, 8 May 2012 21:19:21 +0000 (21:19 +0000)]
[analyzer] SelfInit: Stop tracking self if it's assigned a value we
don't reason about.
Self is just like a local variable in init methods, so it can be
assigned anything like result of static functions, other methods ... So
to suppress false positives that result in such cases, stop tracking the
checker-specific state after self is being assigned to (unless the
value is't being assigned to is either self or conforms to our rules).
This change does not invalidate any existing regression tests.
Gregory Szorc [Tue, 8 May 2012 05:56:38 +0000 (05:56 +0000)]
[clang.py] TranslationUnit API improvements
* TranslationUnit reading and parsing now implemented as
TranslationUnit.from_ast() and TranslationUnit.from_source().
* Index.parse() and Index.read() implemented through above.
* Index.parse() and Index.read() now raise a TanslationUnitLoadException
instead of returning None if a TranslationUnit could not be
instantiated. This is backwards incompatible.
* Ability to save TranslationUnits via TranslationUnit.save().
* TranslationUnit now holds onto Index instance that created. This means
the Index can't be GC'd until the TranslationUnit is itself GC'd,
making memory management thoughtless.
* Don't use [] as a default argument value, as the initial value used is
reused for the duration of the program.
Jordy Rose [Tue, 8 May 2012 03:27:16 +0000 (03:27 +0000)]
[analyzer] Rework both constraint managers to handle mixed-type comparisons.
This involves keeping track of three separate types: the symbol type, the
adjustment type, and the comparison type. For example, in "$x + 5 > 0ULL",
if the type of $x is 'signed char', the adjustment type is 'int' and the
comparison type is 'unsigned long long'. Most of the time these three types
will be the same, but we should still do the right thing when the
comparison value is out of range, and wraparound should be calculated in
the adjustment type.
This also re-disables an out-of-bounds test; we were extracting the symbol
from non-additive SymIntExprs, but then throwing away the integer.
Sorry for the large patch; both the basic and range constraint managers needed
to be updated together, since they share code in SimpleConstraintManager.
Jordy Rose [Tue, 8 May 2012 03:26:58 +0000 (03:26 +0000)]
[analyzer] Add an abstraction for the bit width and signedness of an APSInt. No functionality change.
There are more parts of the analyzer that could use the convenience of APSIntType, particularly the constraint engine, but that needs a fair amount of rewriting to handle mixed-type constraints anyway.
Douglas Gregor [Tue, 8 May 2012 00:14:45 +0000 (00:14 +0000)]
Introduce a new libclang API to determine the platform availability of
a given entity, so that we can tell when the entity was
introduced/deprecated/obsoleted on each platform for which we have an
annotation. Addresses <rdar://problem/11365715>.
Process attributes in the order they appear in the source code. This make clang
match gcc behavior for two conflicting visibilities in the same decl. It also
makes handling of dllimport/dllexport more natural.
[libclang] Use a copy of the data-recursive RecursiveASTVisitor inside libclang.
This is temporary until we settle on how to make the main RecursiveASTVisitor safe.
There are some modifications on the original version, to avoid extreme memory usage
when compiling for release.
Remove the 'intrin.h' builtin header file and its tests for now.
After discussion with several people, including Doug Gregor, we've
decided to change our approach here. If you have questions about this
header file, the commit removing it, etc., please reach out to me
off-list.
Anna Zaks [Mon, 7 May 2012 19:40:49 +0000 (19:40 +0000)]
[analyzer] Fix a crash in triggered by OSAtomicChecker.
SValBuilder should return an UnknownVal() when comparison of int and ptr
fails. Previous to this commit, it went on assuming that we are dealing
with pointer arithmetic.
Anna Zaks [Mon, 7 May 2012 18:40:26 +0000 (18:40 +0000)]
[analyzer] Fixup for r155244: PurgeDeadSymbols should subclass StmtPoint
To solve the inconsistency pointed out in Erik's review, refactor class
hierarchy of ProgramPoints so that PreStmtPurgeDeadSymbols and
PostStmtPurgeDeadSymbols both subclass from StmtPoint instead of
PostStmt.
Nuno Lopes [Mon, 7 May 2012 17:57:00 +0000 (17:57 +0000)]
remove code to add bound checks for simple array accesses, since those are already covered by the check with the objectsize builtin
remove the comparison of objectsize with -1. since it's an unsigned comparison, it will always succeed if objectsize returns -1, which is enough to have the check removed
Manuel Klimek [Mon, 7 May 2012 09:17:48 +0000 (09:17 +0000)]
Fixes resolution of relative paths when running clang tools.
The chdir is not the perfect fix, as it is thread hostile. The
real fix will be to make -working-dir work correctly, which will
take time to implement. Before that, the tooling library cannot
be used concurrently.
Richard Smith [Mon, 7 May 2012 09:03:25 +0000 (09:03 +0000)]
When we suppress an error due to SFINAE, stash the diagnostic away with the
overload candidate, and include its message in any subsequent 'candidate not
viable due to substitution failure' note we may produce.
To keep the note small (since the 'overload resolution failed' diagnostics are
often already very verbose), the text of the SFINAE diagnostic is included as
part of the text of the note, and any notes which were attached to it are
discarded.
There happened to be spare space in OverloadCandidate into which a
PartialDiagnosticAt could be squeezed, and this patch goes to lengths to avoid
unnecessary PartialDiagnostic copies, resulting in no slowdown that I could
measure. (Removal in passing of some PartialDiagnostic copies has resulted in a
slightly smaller clang binary overall.) Even on a torture test, I was unable to
measure a memory increase of above 0.2%.
John McCall [Mon, 7 May 2012 06:16:58 +0000 (06:16 +0000)]
Change how we suppress access control in explicit instantiations
so that we actually accumulate all the delayed diagnostics. Do
this so that we can restore those diagnostics to good standing
if it turns out that we were wrong to suppress, e.g. if the
tag specifier is actually an elaborated type specifier and not
a declaration.
John McCall [Mon, 7 May 2012 06:16:41 +0000 (06:16 +0000)]
Refactor DelayedDiagnostics so that it keeps diagnostics in
separate pools owned by the RAII objects that keep pushing
decl state. This gives us quite a bit more flexibility.
Manuel Klimek [Mon, 7 May 2012 05:56:03 +0000 (05:56 +0000)]
- Adding lexical_parent and semantic_parent properties to clang.cindex.Cursor
- Two new tests (one for each property), require libclang built from r155858 or later to pass
- New test utility function (get_cursors) that gets all the nodes with a specific spelling.
Richard Smith [Mon, 7 May 2012 01:07:30 +0000 (01:07 +0000)]
A union can have a constexpr defaulted default constructor, if it has an
in-class initializer for one of its fields. Value-initialization of such
a type should use the in-class initializer!
The former was just a bug, the latter is a (reported) standard defect.
Jordy Rose [Sun, 6 May 2012 23:40:02 +0000 (23:40 +0000)]
[analyzer] Reduce parallel code paths in SimpleSValBuilder::evalBinOpNN, and handle mixed-type operations more generally.
The logical change is that the integers in SymIntExprs may not have the same type as the symbols they are paired with. This was already the case with taint-propagation expressions created by SValBuilder::makeSymExprValNN, but I think those integers may never have been used. SimpleSValBuilder should be able to handle mixed-integer-type SymIntExprs fine now, though, and the constraint managers were already being defensive (though not entirely correct). All existing tests pass.
The logic in evalBinOpNN has been simplified so that conversion is done as late as possible. As a result, most of the switch cases have been reduced to do the minimal amount of work, delegating to another case when they can by substituting ConcreteInts and (as before) reversing the left and right arguments when useful.
Comparisons require special handling in two places (building SymIntExprs and evaluating constant-constant operations) because we don't /know/ the best type for comparing the two values. I've approximated the rules in Sema [C99 6.3.1.8] but it'd be nice to refactor Sema's actual algorithm into ASTContext.
This is also groundwork for handling mixed-type constraints better than we do now.
Split mergeAvailabilityAttr out of handleAvailabilityAttr. This is important
for having a uniform logic for adding attributes to a decl. This in turn
is needed to fix the FIXME:
// FIXME: This needs to happen before we merge declarations. Then,
// let attribute merging cope with attribute conflicts.
ProcessDeclAttributes(S, NewFD, D,
/*NonInheritable=*/false, /*Inheritable=*/true);
The idea is that mergeAvailabilityAttr will become a method. Once attributes
are processed before merging, it will be called from handleAvailabilityAttr to
handle multiple attributes in one decl:
NAKAMURA Takumi [Sun, 6 May 2012 08:24:34 +0000 (08:24 +0000)]
Revert r156142, "[libclang] Bring back BodyIndexer::TraverseCXXOperatorCallExpr which"
It caused test/Index/index-many-call-ops.cpp to fail in stage2 c-index-test on selfhosting i686-cygwin and x86_64-linux since r156229 (Reverting making RecursiveASTVisitor data recursive).
Richard Smith [Sun, 6 May 2012 00:04:32 +0000 (00:04 +0000)]
A conversion operator in a base class shouldn't hide another conversion operator
in the same class, even if they convert to the same type. Fixes PR12712.
Ted Kremenek [Sat, 5 May 2012 04:35:20 +0000 (04:35 +0000)]
Revert r156141 (making RecursiveASTVisitor data recursive). It is causing clang to blow up in memory usage on -O2 when compiling itself,
which is leading to swapping in some cases when it didn't before. We need to see if we can make this change
without leading to a massive compile-time bloat.
Douglas Gregor [Fri, 4 May 2012 22:38:52 +0000 (22:38 +0000)]
Move Sema::VerifyIntegerConstantExpression() and
Sema::ConvertToIntegralOrEnumerationType() from PartialDiagnostics to
abstract "diagnoser" classes. Not much of a win here, but we're
-several PartialDiagnostics.
Douglas Gregor [Fri, 4 May 2012 18:24:37 +0000 (18:24 +0000)]
Synchronize the representations of DeclarationName and Selector so
that bridging between the two is free. Saves ~4k of code size,
although I don't see any measurable performance difference
(unfortunately).
Sean Callanan [Fri, 4 May 2012 18:22:53 +0000 (18:22 +0000)]
IsTailPaddedMemberArray uses a FieldDecl's
getTypeSourceInfo() without checking for NULL.
FieldDecls may have NULL TypeSourceInfo, and in
fact some FieldDecls generated by Clang -- and
all FieldDecls generated by LLDB -- have no
TypeSourceInfo.
This patch makes IsTailPaddedMemberArray check
for NULL.
Douglas Gregor [Fri, 4 May 2012 16:32:21 +0000 (16:32 +0000)]
Move Sema::RequireCompleteType() and Sema::RequireCompleteExprType()
off PartialDiagnostic. PartialDiagnostic is rather heavyweight for
something that is in the critical path and is rarely used. So, switch
over to an abstract-class-based callback mechanism that delays most of
the work until a diagnostic is actually produced. Good for ~11k code
size reduction in the compiler and 1% speedup in -fsyntax-only on the
code in <rdar://problem/11004361>.
James Molloy [Fri, 4 May 2012 10:55:22 +0000 (10:55 +0000)]
Fix handling of wint_t - we can't assume wint_t is purely an integer promotion of wchar_t - they may differ in signedness.
Teach ASTContext about WIntType, and have it taken from TargetInfo like WCharType. Should fix test/Sema/format-strings.c for ARM, with the exception of one subtest which will fail if wint_t and wchar_t are the same size and wint_t is signed, wchar_t is unsigned.
There'll be a followup commit to fix that.
Reviewed by Chandler and Hans at http://llvm.org/reviews/r/8
Alexey Samsonov [Fri, 4 May 2012 07:39:27 +0000 (07:39 +0000)]
This patch adds a new Clang compiler flag "-gline-tables-only".
It reduces the amount of emitted debug information:
1) DIEs in .debug_info have types DW_TAG_compile_unit, DW_TAG_subprogram,
DW_TAG_inlined_subroutine (for opt builds) and DW_TAG_lexical_block only.
2) .debug_str contains only function names.
3) No debug data for types/namespaces/variables is emitted.
4) The data in .debug_line is enough to produce valid stack traces with
function names and line numbers.
[libclang] Bring back BodyIndexer::TraverseCXXOperatorCallExpr which
was removed in r155969 to address a deficiency of RecursiveASTVisitor
prior to recent changes on it.
Make RecursiveASTVisitor fully data recursive on Stmts, with
minimal disruption on its clients.
Unlike the previous data-recursive scheme, Traverse*Stmt methods are
always getting called. The base methods of RecursiveASTVisitor will enqueue
the sub-statements instead of calling TraverseStmt on them.
Clients that override a Traverse*Stmt method and call TraverseStmt will
still function as function-recursive traversal; if a client wants to
enqueue a sub-statement in its override method it can do it like this:
[inside the override method]
StmtQueueAction StmtQueue(*this);
StmtQueue.queue(Stmt->getSubStmt());
Fix non-deterministic iteration order when walking the specializations
of templates by using the newly introduce FoldingSetVector. This
preserves insertion order for all iteration of specializations.
I've also included a somewhat terrifying testcase that rapidly builds up
a large number of functions. This is enough that any system with ASLR
will have non-deterministic debug information generated for the test
case without the fix here as the debug information is generated in part
by walking these specializations.
Douglas Gregor [Thu, 3 May 2012 23:18:44 +0000 (23:18 +0000)]
Split DeclarationName::getFETokenInfoAsVoid() into hot/cold paths and
(trivially) make DeclContext::lookup()'s const version inlinable. Good
for 0.3% on <rdar://problem/11004361>.
Chad Rosier [Thu, 3 May 2012 22:38:00 +0000 (22:38 +0000)]
[driver - crash diagnostics] Convert the flags back to an array of strings and use
array_lengthof. Also, append the new filename with correct preprocessed suffix.
Last part of rdar://11285725
Richard Smith [Thu, 3 May 2012 22:18:20 +0000 (22:18 +0000)]
Use the standard values for the __cplusplus macro, even when in GNU mode. GCC
dropped its prior behavior of always defining __cplusplus to 1 in GNU mode in
version 4.7.
Douglas Gregor [Thu, 3 May 2012 21:25:24 +0000 (21:25 +0000)]
Do not keep track of the set of visited Objective-C containers when
performing the search for overridden methods. We very rarely see the
same container twice, and in those rare cases we still have the
fallback of the second SmallPtrSet to eliminate duplicates. Good for
~1.5% -fsyntax-only speedup on the code in <rdar://problem/11004361>.
projects / clang / log