Richard Smith [Mon, 29 Oct 2012 18:26:47 +0000 (18:26 +0000)]
Partially roll back r166898; it exposed a bug in the standard.
The problem is as follows: C++11 has contexts which are not
potentially-evaluated, and yet in which we are required or encouraged to
perform constant evaluation. In such contexts, we are not permitted to
implicitly define special member functions for literal types, therefore
we cannot evalaute those constant expressions.
Punt on this in one more context for now by skipping checking constexpr
variable initializers if they occur in dependent contexts.
Jordan Rose [Mon, 29 Oct 2012 17:46:47 +0000 (17:46 +0000)]
-Warc-repeated-use-of-weak: allow single reads in loops from local variables.
Previously, the warning would erroneously fire on this:
for (Test *a in someArray)
use(a.weakProp);
...because it looks like the same property is being accessed over and over.
However, clearly this is not the case. We now ignore loops like this for
local variables, but continue to warn if the base object is a parameter,
global variable, or instance variable, on the assumption that these are
not repeatedly usually assigned to within loops.
Additionally, do-while loops where the condition is 'false' are not really
loops at all; usually they're just used for semicolon-swallowing macros or
using "break" like "goto".
Jordan Rose [Mon, 29 Oct 2012 17:31:59 +0000 (17:31 +0000)]
[analyzer] New option to not suppress null return paths if an argument is null.
Our one basic suppression heuristic is to assume that functions do not
usually return NULL. However, when one of the arguments is NULL it is
suddenly much more likely that NULL is a valid return value. In this case,
we don't suppress the report here, but we do attach /another/ visitor to
go find out if this NULL argument also comes from an inlined function's
error path.
This new behavior, controlled by the 'avoid-suppressing-null-argument-paths'
analyzer-config option, is turned off by default. Turning it on produced
two false positives and no new true positives when running over LLVM/Clang.
This is one of the possible refinements to our suppression heuristics.
<rdar://problem/12350829>
Jordan Rose [Mon, 29 Oct 2012 17:31:53 +0000 (17:31 +0000)]
[analyzer] Use the CallEnter node to get a value for tracked null arguments.
Additionally, don't collect PostStore nodes -- they are often used in
path diagnostics.
Previously, we tried to track null arguments in the same way as any other
null values, but in many cases the necessary nodes had already been
collected (a memory optimization in ExplodedGraph). Now, we fall back to
using the value of the argument at the time of the call, which may not
always match the actual contents of the region, but often will.
This is a precursor to improving our suppression heuristic.
<rdar://problem/12350829>
Bill Schmidt [Mon, 29 Oct 2012 14:59:24 +0000 (14:59 +0000)]
This patch adds alignment information for long double to the 64-bit PowerPC
ELF subtarget.
The existing description string is moved from PPC64TargetInfo to its
DarwinTargetInfo subclass, to avoid any changes to the Darwin ABI.
PPC64TargetInfo now has two possible description strings: one for FreeBSD,
which requires 8-byte alignment, and a default string that requires
16-byte alignment.
I've added a test for PPC64 Linux to verify the 16-byte alignment. If
somebody wants to add a separate test for FreeBSD, that would be great.
Note that there is a companion patch to update the alignment information
in LLVM, which I am committing now as well.
Richard Smith [Sun, 28 Oct 2012 07:39:29 +0000 (07:39 +0000)]
Revert functional part of r166896 and just suppress -Wunneeded-internal-declaration for reference types for now. This needs more work; the cases we currently miss are a bit random.
Richard Smith [Sun, 28 Oct 2012 06:18:02 +0000 (06:18 +0000)]
When determining whether to try evaluating the initializer of a variable, check
whether the initializer is value-dependent rather than whether we are in a
dependent context. This allows us to detect some errors sooner, and fixes a
crash-on-invalid if a dependent type leaks out to a non-dependent context in
error recovery.
Richard Smith [Sun, 28 Oct 2012 04:47:21 +0000 (04:47 +0000)]
In -Wunneeded-internal-declaration, suppress the warning for variables which
might have been used in constant expressions, rather than suppressing it for
variables which are const. The important thing here is that such variables
can have their values used without actually being marked as 'used'.
In Parser::ParseDecltypeSpecifier, make sure the end location it returns
is at the end of parsed tokens when an error occurs, otherwise we'll hit
an assertion when trying to annotate the decltype tokens.
Douglas Gregor [Fri, 26 Oct 2012 22:31:14 +0000 (22:31 +0000)]
When an externally-supplied record layout has a size that clearly
doesn't include padding up to the alignment of the record, take this
as a cue that the alignment of the record should (conservatively) be
set to 1. This is similar to other the other cues we use to determine
that the record has a lower alignment, e.g., that the
externally-supplied layout places fields at lower offsets than we
would. Fixes <rdar://problem/12582052>; test case in LLDB.
objective-C IRGen: for @implementation nested in
extern "C", its method definitions must be IRGen'ed
before meta-data for class is generated. Otherwise,
IRGen crashes (to say the least).
// rdar://12581683
Bill Schmidt [Fri, 26 Oct 2012 19:59:03 +0000 (19:59 +0000)]
This patch addresses a 64-bit PowerPC ELF ABI compatibility issue with
varargs parameter passing.
A strict reading of the ABI indicates that any argument with alignment greater
than 8 may require skipping doublewords in the parameter save area to align
the argument, and hence require skipping GPRs. In practice, this is not done
by GCC. The alignment restriction is used for internal alignment of a
structure, but a structure with 16-byte alignment, for example, is not
itself 16-byte aligned in the parameter save area. Although this is messy,
it has become the de facto standard used in building existing libraries.
My initial varargs support followed the ABI language, but not the de facto
standard. Running the GCC compatibility test suite exposed this issue, and
indeed showed that LLVM didn't pass parameters self-consistently with my
original logic. Removing the additional alignment logic allows the affected
tests to now pass.
I modified the ppc64-varargs-struct.c test case to remove the existing test
for generation of alignment code, which is no longer appropriate.
Built and tested on powerpc64-unknown-linux-gnu with no new regressions.
Douglas Gregor [Fri, 26 Oct 2012 16:45:11 +0000 (16:45 +0000)]
Match up anonymous structs/unions in the ASTImporter. Previously, we'd
only actually get the answer right if there was only a single
anonymous struct/union at that level. This is part of
<rdar://problem/11904570>; the test will go into LLDB itself.
Bill Wendling [Thu, 25 Oct 2012 23:28:48 +0000 (23:28 +0000)]
Recommit Eric's code to validate ASM string's constraints and modifiers.
This code checks the ASM string to see if the output size is able to fit within
the variable specified as the output. For instance, scalar-to-vector conversions
may not really work. It's on by default, but can be turned off with a flag if
you think you know what you're doing.
This is placed under a flag ('-Wasm-operand-widths') and flag group ('-Wasm').
Richard Smith [Thu, 25 Oct 2012 22:27:30 +0000 (22:27 +0000)]
LLVM's hashing routines produce a size_t, and thus generate different values for 32- and 64-bit host compilers. This really needs to be fixed -- the IR generated should not depend on the host -- but this change will get the bots green again. Proper fix to follow.
Ted Kremenek [Thu, 25 Oct 2012 22:07:10 +0000 (22:07 +0000)]
TrackConstraintBRVisitor and ConditionBRVisitor can emit similar
path notes for cases where a value may be assumed to be null, etc.
Instead of having redundant diagnostics, do a pass over the generated
PathDiagnostic pieces and remove notes from TrackConstraintBRVisitor
that are already covered by ConditionBRVisitor, whose notes tend
to be better.
David Tweed [Thu, 25 Oct 2012 13:56:30 +0000 (13:56 +0000)]
Correct test inovocations to use %clang_cc1 rather than direct invocation (so that it can have additional options set when trying to debug issues causing regressions).
Alexey Samsonov [Thu, 25 Oct 2012 10:18:50 +0000 (10:18 +0000)]
Initialize debug info for special cases of functions that lack declarations and are generated by Clang (global initializers/destructors, thunks) . Fixes PR13942.
Douglas Gregor [Thu, 25 Oct 2012 00:30:23 +0000 (00:30 +0000)]
Remove the old predefines-buffer diffing code completely. It's been
replaced by the more efficient, cleaner preprocessor-option version
that occurs earlier in PCH validation.