Rafael Espindola [Fri, 18 May 2012 01:47:00 +0000 (01:47 +0000)]
Centralize the handling of the "attribute declaration must precede definition"
warning. This also makes us warn on tags, which, ironically, is the only case
gcc warns on.
Douglas Gregor [Thu, 17 May 2012 23:13:29 +0000 (23:13 +0000)]
A selector match between two Objective-C methods does *not* guarantee
that the methods have the same number of parameters, although we
certainly assumed this in many places. Objective-C can be insane
sometimes. Fixes <rdar://problem/11460990>.
Douglas Gregor [Thu, 17 May 2012 22:39:14 +0000 (22:39 +0000)]
In the override search for Objective-C methods, protect against ASTs that have NULL interfaces behind a category, which can happen in invalid code. Fixes <rdar://problem/11478173>, a recent regression
[arcmt] Remove the "it is not safe to remove an unused 'autorelease' message" ARC
migration error.
This is more trouble that it is worth; autoreleasing a value without holding on it
is a valid use-case, we should not "punish" correct code for the minority of
broken/fragile programs that depend on the behavior of -autorelease.
Use the argument location instead of the format string location when warning
about argument type mismatch.
This gives a nicer diagnostic in cases like
printf(fmt,
i);
where previously the snippet just pointed at 'fmt' (with a note at the
definition of fmt).
It's a wash for cases like
printf("%f",
i);
where previously we snippeted the offending portion of the format string,
but didn't indicate which argument was at fault.
[libclang/AST] When declaring a local class, don't neglect to set the end location
of the DeclStmt node, otherwise libclang will not work for anything inside that
class.
Chad Rosier [Wed, 16 May 2012 23:45:12 +0000 (23:45 +0000)]
[driver] Remove obsolete support for -A link option. The standard -A option is
used by the preprocessor. Apple's GCC also supported a -A option for linking.
The ld man page has the following:
-A basefile - Obsolete incremental load format. This option is obsolete.
Nick Kledzik confirms this option is no longer needed/supported.
rdar://11455614
Richard Smith [Wed, 16 May 2012 23:40:17 +0000 (23:40 +0000)]
Recover better from a missing 'typename' in a function template definition.
Disambiguate past such a potential problem, and use the absence of 'typename'
to break ties in favor of a parenthesized thingy being an initializer, if
nothing else in the declaration disambiguates it as declaring a function.
Chad Rosier [Wed, 16 May 2012 20:40:09 +0000 (20:40 +0000)]
[driver] Allow the driver to directly accept the -no-implicit-float option, so that the
generation of implicit floating point instructions can be disable for ARM.
rdar://11409142
Chad Rosier [Wed, 16 May 2012 19:28:02 +0000 (19:28 +0000)]
Warn about -Wno-foo where foo is an unknown warning option. This is helpful
for subtle misspellings such as -Wno-unused-command-line-arguments instead of
-Wno-unused-command-line-argument.
Also fix the diagnostic messages to properly handle -Wno- options. Previously,
the positive version was always emitted (i.e., -Wfoo was emitted for -Wno-foo).
rdar://11461500
Richard Trieu [Wed, 16 May 2012 19:04:59 +0000 (19:04 +0000)]
Move the warnings for extra semi-colons under -Wextra-semi. Also, added
a warning for an extra semi-colon after function definitions. Added logic
so that a block of semi-colons on a line will only get one warning instead
of a warning for each semi-colon.
Douglas Gregor [Wed, 16 May 2012 17:01:33 +0000 (17:01 +0000)]
Clean up r156925, so that we only mark the capturing DeclRefExpr of a
lambda as referring to a local in an enclosing scope if we're in the
enclosing scope of the lambda (not it's function call operator). Also,
turn the test into an IR generation test, since that's where the
crashes occurred. Really fixes PR12746 / <rdar://problem/11465120>.
Douglas Gregor [Wed, 16 May 2012 16:50:20 +0000 (16:50 +0000)]
Fix code generation of variables reference expressions when mixing
blocks and lambdas, based heavily on a patch from Meador Inge. Fixes
PR12746 / <rdar://problem/11465120>.
Jordy Rose [Wed, 16 May 2012 16:01:07 +0000 (16:01 +0000)]
[analyzer] Introduce clang_analyzer_eval for regression test constraint checks.
The new debug.ExprInspection checker looks for calls to clang_analyzer_eval,
and emits a warning of TRUE, FALSE, or UNKNOWN (or UNDEFINED) based on the
constrained value of its (boolean) argument. It does not modify the analysis
state though the conditions tested can result in branches (e.g. through the
use of short-circuit operators).
David Blaikie [Wed, 16 May 2012 04:20:04 +0000 (04:20 +0000)]
Include the correct conversion context locations for condition expressions.
This improves the conversion diagnostics (by correctly pointing to the loop
construct for conversions that may've been caused by the contextual conversion
to bool caused by a condition expression) and also causes the NULL conversion
warnings to be correctly suppressed when crossing a macro boundary in such a
context. (previously, since the conversion context location was incorrect, the
suppression could not be performed)
objective-c: perform strict type checking on property
type and its accessor type and issue error if types
are incompatible, instead of crashing in IRgen.
// rdar://1105153
David Blaikie [Tue, 15 May 2012 21:57:38 +0000 (21:57 +0000)]
Don't warn when NULL is used within a macro but its conversion is outside a macro.
This fixes the included test case & was reported by Nico Weber.
It's a little bit nasty using the difference in the conversion context, but
seems to me like a not unreasonable solution. I did have to fix up the
conversion context for conditional operators (it seems correct to me to include
the context for which we're actually doing the comparison - across all the
nested conditionals, rather than the innermost conditional which might not
actually have the problematic implicit conversion at all) and template default
arguments (this is a bit of a hack, since we don't have the source location of
the '=' anymore, so I just used the start of the parameter - open to
suggestions there)
Richard Smith [Tue, 15 May 2012 21:29:55 +0000 (21:29 +0000)]
If we see a declaration which is either missing a type or has a malformed type,
and the thing we have has a scope specifier, and we're in a context that doesn't
allow declaring a qualified name, then the error is a malformed type, not a
missing type.
[objcmt] Rewrite [NSNumber numberWith*] messages to the new @() boxing syntax.
There are some caveats:
-If an implicit cast (e.g. int -> float for numberWithFloat:) was required, the message
will not get rewritten
-If the message was with numberWithInteger:/numberWithUnsignedInteger:, which are very
commonly used, be more liberal and allow the boxing syntax if the underlying type has
same signedness and will not lose precision.
objective-c nonfragile abi: discourage ivar declarations
in @interface by issuing warning (off by default) under
opt'ed in flag -Winterface-block-ivar. // rdar://10763173
Richard Smith [Tue, 15 May 2012 06:21:54 +0000 (06:21 +0000)]
Further improvement to wording of overload resolution diagnostics, and including
the sole parameter name in the diagnostic in more cases. Patch by Terry Long!
Richard Smith [Tue, 15 May 2012 06:15:11 +0000 (06:15 +0000)]
PR12798: Don't drop part of the nested name specifier when instantiating a
pseudo-destructor expression. This can affect whether virtual dispatch for
the destructor call is bypassed.
Richard Smith [Tue, 15 May 2012 04:39:51 +0000 (04:39 +0000)]
Fold the six functions checking explicitly-defaulted special member functions
into one. These were all performing almost identical checks, with different bugs
in each of them.
This fixes PR12806 (we weren't setting the exception specification for an
explicitly-defaulted, non-user-provided default constructor) and enforces
8.4.2/2's rule that an in-class defaulted member must exactly match the implicit
parameter type.
John McCall [Tue, 15 May 2012 02:01:59 +0000 (02:01 +0000)]
Change the mangling of a ref-qualifier on a function type so that
it is placed in a position which is never ambiguous with a
reference-to-function type. This follows some recent discussion
and ensuing proposal on cxx-abi-dev. It is not necessary to
change the mangling of CV-qualifiers because you cannot
apply CV-qualification in the normal sense to a function type.
It is not necessary to change the mangling of ref-qualifiers on
method declarations because they appear in an unambiguous
location.
In addition, mangle CV-qualifiers and ref-qualifiers on function
types when they occur in positions other than member pointers
(that is, when they appear as template arguments).
This is a minor ABI break with previous releases of clang. It
is not considered critical because (1) ref-qualifiers are
relatively rare, since AFAIK we're the only implementing compiler,
and (2) they're particularly likely to come up in contexts that
do not rely on the ODR for correctness. We apologize for any
inconvenience; this is the right thing to do.
Richard Smith [Mon, 14 May 2012 22:43:34 +0000 (22:43 +0000)]
Recover properly from a redundant 'typename' before a non-nested name. This is
permitted as a Microsoft extension. Patch by William Wilson! (Plus some minor
tweaking by me.)
Jordy Rose [Mon, 14 May 2012 17:58:35 +0000 (17:58 +0000)]
[analyzer] strncpy: Special-case a length of 0 to avoid an incorrect warning.
We check the address of the last element accessed, but with 0 calculating that
address results in element -1. This patch bails out early (and avoids a bunch
of other work at that).
projects / clang / log