Richard Smith [Sun, 21 Apr 2013 01:08:50 +0000 (01:08 +0000)]
The 'constexpr implies const' rule for non-static member functions is gone in
C++1y, so stop adding the 'const' there. Provide a compatibility warning for
code relying on this in C++11, with a fix-it hint. Update our lazily-written
tests to add the const, except for those ones which were testing our
implementation of this rule.
Richard Smith [Sat, 20 Apr 2013 23:28:26 +0000 (23:28 +0000)]
Disable VLA diagnostic in C++1y mode, and add some tests.
Still to do here:
- we have a collection of syntactic accepts-invalids to diagnose
- support non-PODs in VLAs, including dynamic initialization /
destruction
- runtime checks (and throw std::bad_array_length) for bad bound
- support VLA capture by reference in lambdas
- properly support VLAs in range-based for (don't recompute bound)
Richard Smith [Sat, 20 Apr 2013 22:23:05 +0000 (22:23 +0000)]
C++1y: Allow aggregates to have default initializers.
Add a CXXDefaultInitExpr, analogous to CXXDefaultArgExpr, and use it both in
CXXCtorInitializers and in InitListExprs to represent a default initializer.
There's an additional complication here: because the default initializer can
refer to the initialized object via its 'this' pointer, we need to make sure
that 'this' points to the right thing within the evaluation.
Richard Smith [Sat, 20 Apr 2013 12:57:49 +0000 (12:57 +0000)]
Clarifying memory allocation: approved for C++14. Move from N/A to no, since we currently relax 'operator new' calls which didn't come from new-expressions.
Sourcery CodeBench and modern FSF Mips toolchains require a bit more
complicated algorithm to calculate headers, libraries and sysroot paths
than implemented by Clang driver now. The main problem is that all these
paths depend on a set of command line arguments additionally to a target
triple value. For example, let $TC is a toolchain installation directory.
If we compile big-endian 32-bit mips code, crtbegin.o is in the
$TC/lib/gcc/mips-linux-gnu/4.7.2 folder and the toolchain's linker requires
--sysroot=$TC/mips-linux-gnu/libc argument. If we compile little-endian
32-bit soft-float mips code, crtbegin.o is in the
$TC/lib/gcc/mips-linux-gnu/4.7.2/soft-float/el folder and the toolchain's
linker requires --sysroot=$TC/mips-linux-gnu/libc/soft-float/el argument.
1. Calculate MultiarchSuffix using all necessary command line options and
use this MultiarchSuffix to detect crtbegin.o location in the
GCCInstallationDetector::ScanLibDirForGCCTriple() routine.
2. If a user does not provide --sysroot argument to the driver explicitly,
calculate new sysroot value based on command line options. Then use this
calculated sysroot path:
a. To populate a file search paths list in the Linux::Linux() constructor.
b. To find Mips toolchain specific include headers directories
in the Linux::AddClangSystemIncludeArgs() routine.
c. To provide -–sysroot argument for a linker.
Note:
- The FSF's tree slightly differs (folder names) and is not supported
yet.
- New addExternCSystemIncludeIfExits() routine is a temporary solution.
I plan to move path existence check to the addExternCSystemInclude()
routine by a separate commit.
The patch reviewed by Rafael Espindola.
http://llvm-reviews.chandlerc.com/D644
Anna Zaks [Sat, 20 Apr 2013 01:15:42 +0000 (01:15 +0000)]
[analyzer] Ensure BugReporterTracking works on regions with pointer arithmetic
Introduce a new helper function, which computes the first symbolic region in
the base region chain. The corresponding symbol has been used for assuming that
a pointer is null. Now, it will also be used for checking if it is null.
This ensures that we are tracking a null pointer correctly in the BugReporter.
Jordan Rose [Fri, 19 Apr 2013 22:19:14 +0000 (22:19 +0000)]
[analyzer] Website: update lists of potential and actual checkers.
- memory.MismatchedDelete, memory.MultipleDelete, and memory.DeallocateNonPtr
are complete (unix.MismatchedDeallocator and cplusplus.NewDelete)
- Per discussion on the mailing list, different.UnaryPlusWithUnsigned has
dubious value; remove it.
- Add potential checker ctordtor.PlacementSelfCopy per an internal bug report.
- core.AttributeNonNull is now core.NonNullParamChecker, though no one should
be depending on this name anyway.
Chad Rosier [Fri, 19 Apr 2013 20:37:49 +0000 (20:37 +0000)]
[ms-inline asm] The parsing of C++ identifiers is a task of the front-end parser,
not the asm parser. As such, begin moving the parsing logic in that direction.
This patch is just a temporary hack until the real frontend parser can be hooked
up. Part of rdar://13663589
Avoid names like __in that conflict with SAL in builtin headers
Microsoft's Source Annotation Language (SAL) defines a bunch of keywords
for annotating the inputs and outputs of functions. Empty definitions
for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>.
This makes it basically impossible to include MSVC's stdlib.h and
Clang's *mmintrin.h headers at the same time if they have variables
named __in. As a workaround, I've renamed those variables.
This fixes the Modules/compiler_builtins.m test which was XFAILed,
presumably due to this conflict.
[libclang] Introduce clang_Cursor_getObjCDeclQualifiers, to query for 'ObjC Qualifiers' written next to the return and
parameter types in an ObjC method declarations.
Andy Gibbs [Thu, 18 Apr 2013 16:49:37 +0000 (16:49 +0000)]
Dropped the parentheses for #pragma message and its kin in the -E output generator.
This was a suggestion by Jordan Rose since the documented format for these pragmas is without the parentheses. At the same time, I've increased test coverage too for the preprocessed output.
Jordan Rose [Thu, 18 Apr 2013 16:33:46 +0000 (16:33 +0000)]
[analyzer] "Force" LazyCompoundVals on bind when they are simple enough.
The analyzer uses LazyCompoundVals to represent rvalues of aggregate types,
most importantly structs and arrays. This allows us to efficiently copy
around an entire struct, rather than doing a memberwise load every time a
struct rvalue is encountered. This can also keep memory usage down by
allowing several structs to "share" the same snapshotted bindings.
However, /lookup/ through LazyCompoundVals can be expensive, especially
since they can end up chaining back to the original value. While we try
to reuse LazyCompoundVals whenever it's safe, and cache information about
this transitivity, the fact is it's sometimes just not a good idea to
perpetuate LazyCompoundVals -- the tradeoffs just aren't worth it.
This commit changes RegionStore so that binding a LazyCompoundVal to struct
will do a memberwise copy if the struct is simple enough. Today's definition
of "simple enough" is "up to N scalar members" (see below), but that could
easily be changed in the future. This is enough to bring the test case in
PR15697 back down to a manageable analysis time (within 20% of its original
time, in an unfair test where the new analyzer is not compiled with LTO).
The actual value of "N" is controlled by a new -analyzer-config option,
'region-store-small-struct-limit'. It defaults to "2", meaning structs with
zero, one, or two scalar members will be considered "simple enough" for
this code path.
It's worth noting that a more straightforward implementation would do this
on load, not on bind, and make use of the structure we already have for this:
CompoundVal. A long time ago, this was actually how RegionStore modeled
aggregate-to-aggregate copies, but today it's only used for compound literals.
Unfortunately, it seems that we've special-cased LazyCompoundVal in certain
places (such as liveness checks) but failed to similarly special-case
CompoundVal in all of them. Until we're confident that CompoundVal is
handled properly everywhere, this solution is safer, since the entire
optimization is just an implementation detail of RegionStore.
Jordan Rose [Thu, 18 Apr 2013 16:33:40 +0000 (16:33 +0000)]
[analyzer] Don't crash if we cache out after making a temporary region.
A C++ overloaded operator may be implemented as an instance method, and
that instance method may be called on an rvalue object, which has no
associated region. The analyzer handles this by creating a temporary region
just for the evaluation of this call; however, it is possible that /by
creating the region/, the analyzer ends up in a previously-explored state.
In this case we don't need to continue along this path.
This doesn't actually show any behavioral change now, but it starts being
used with the next commit and prevents an assertion failure there.
Richard Trieu [Thu, 18 Apr 2013 01:04:37 +0000 (01:04 +0000)]
Switch the note order for -Woverloaded-shift-op-parentheses so that the note
with the silence fix-it comes first. This is more consistent with the rest
of the warnings in -Wparentheses.
Anna Zaks [Thu, 18 Apr 2013 00:15:15 +0000 (00:15 +0000)]
[analyzer] Tweak getDerefExpr more to track DeclRefExprs to references.
In the committed example, we now see a note that tells us when the pointer
was assumed to be null.
This is the only case in which getDerefExpr returned null (failed to get
the dereferenced expr) throughout our regression tests. (There were multiple
occurrences of this one.)
Anna Zaks [Wed, 17 Apr 2013 22:29:47 +0000 (22:29 +0000)]
[analyzer] Allow TrackConstraintBRVisitor to work when the value it’s tracking is not live in the last node of the path
We always register the visitor on a node in which the value we are tracking is live and constrained. However,
the visitation can restart at a node, later on the path, in which the value is under constrained because
it is no longer live. Previously, we just silently stopped tracking in that case.
[Modules] Use global index to improve typo correction performance
Typo correction for an unqualified name needs to walk through all of the identifier tables of all modules.
When we have a global index, just walk its identifier table only.
[document parsing]: When tag declaration (but not definition!)
is part of the decl-specifier-seq of some other declaration,
it doesn't get comment. // rdar://12390371
Jordan Rose [Wed, 17 Apr 2013 19:09:18 +0000 (19:09 +0000)]
Fix off-by-one error in #pragma clang system_header.
The system_header pragma (from GCC) is implemented using line notes in the
source manager. However, a line note's line number specifies the number
not for the current line, but for the next line. This was making all
line numbers appear off by one after the pragma.
Jordan Rose [Wed, 17 Apr 2013 18:03:48 +0000 (18:03 +0000)]
[analyzer] Don't warn for returning void expressions in void blocks.
This was slightly tricky because BlockDecls don't currently store an
inferred return type. However, we can rely on the fact that blocks with
inferred return types will have return statements that match the inferred
type.
Unified token breaking logic: support for line comments.
Summary:
Added BreakableLineComment, moved common code from
BreakableBlockComment to newly added BreakableComment. As a side-effect of the
rewrite, found another problem with escaped newlines and had to change
code which removes trailing whitespace from line comments not to break after
this patch.
Also,
- abstract out the indirect/in memory/in registers decisions into the CGCXXABI
- fix handling of empty struct arguments for '-cxx-abi microsoft'
- add/fix tests
Andy Gibbs [Wed, 17 Apr 2013 08:06:46 +0000 (08:06 +0000)]
Extended VerifyDiagnosticConsumer to also verify source file for diagnostic.
VerifyDiagnosticConsumer previously would not check that the diagnostic and
its matching directive referenced the same source file. Common practice was
to create directives that referenced other files but only by line number,
and this led to problems such as when the file containing the directive
didn't have enough lines to match the location of the diagnostic in the
other file, leading to bizarre file formatting and other oddities.
This patch causes VerifyDiagnosticConsumer to match source files as well as
line numbers. Therefore, a new syntax is made available for directives, for
example:
This extends the @line feature where "file" is the file where the diagnostic
is generated. The @line syntax is still available and uses the current file
for the diagnostic. "file" can be specified either as a relative or absolute
path - although the latter has less usefulness, I think! The #include search
paths will be used to locate the file and if it is not found an error will be
generated.
The new check is not optional: if the directive is in a different file to the
diagnostic, the file must be specified. Therefore, a number of test-cases
have been updated with regard to this.
Eric Christopher [Wed, 17 Apr 2013 07:19:56 +0000 (07:19 +0000)]
Add a bit of a hack to deal with a failing testcase on darwin10 bots.
We currently emit an error message when you try to use thread local
storage on targets that don't support it and testing C++11 thread
locals will trip this. We don't want to xfail the test for all darwin
hosts so add a quick hack to check for darwin10 and disable the
test based on that. Only checking darwin10 because anything earlier
is really old and I don't have a list of what other hosts don't
support tls handy.
Jordan Rose [Wed, 17 Apr 2013 00:57:39 +0000 (00:57 +0000)]
[analyzer] Merge C++ status page into Open Projects.
Also, add a few random extra open projects.
Most of C++ support is done; we don't need the status page anymore. We're
hoping that the C++-related open projects are the only major pieces of
functionality we don't model at this point.
Use the extra info in global method pool to speed up looking for ObjC overridden methods.
When we are in a implementation, we check the global method pool whether there were category
methods with the same selector. If there were none (common case) we don't need to do lookups for
overridden methods again.
Note that for an interface method (if we don't encounter its implementation), it is considered that
it overrides methods that were declared before it, not for category methods introduced after it.
This is tradeoff in favor of performance, since it is expensive to do lookups in case there was a
category, and moving the global method pool to ASTContext (so we can check it) would increase complexity.
Enhance the ObjC global method pool to record whether there were 0, 1, or >= 2 methods (with a particular selector) inside categories.
This is done by extending ObjCMethodList (which is only used by the global method pool) to have 2 extra bits of information.
We will later take advantage of this info in global method pool for the overridden methods calculation.
projects / clang / log