Rafael Espindola [Thu, 14 Feb 2013 01:47:04 +0000 (01:47 +0000)]
merge hasCLanguageLinkage and isExternC. Keep the shorter name.
I added hasCLanguageLinkage while fixing some language linkage bugs some
time ago so that I wouldn't have to check all users of isExternC. It turned
out to be a much longer detour than expected, but this patch finally
merges the two again. The isExternC function now implements just the
standard notion of having C language linkage.
Rafael Espindola [Thu, 14 Feb 2013 01:18:37 +0000 (01:18 +0000)]
Add a getLanguageLinkage method to VarDecls and FunctionDecls. Use it to fix
some cases where functions with no language linkage were being treated as having
C language linkage. In particular, don't warn in
extern "C" {
static NonPod foo();
}
Since getLanguageLinkage checks the language linkage, the linkage computation
cannot use the language linkage. Break the loop by checking just the context
in the linkage computation.
Nick Lewycky [Thu, 14 Feb 2013 00:55:17 +0000 (00:55 +0000)]
When marking derived classes' virtual methods ODR-used in order to trigger
instantiation in order to permit devirtualization later in codegen, skip over
pure functions since those can't be devirtualization targets.
objective-C: Make order of ivars which are synthesized
in the course of property synthesis deterministic (ordered
by their type size), instead of having hashtable order
(as it is currently). // rdar://13192366
Richard Smith [Wed, 13 Feb 2013 21:18:23 +0000 (21:18 +0000)]
ubsan: Add checking for invalid downcasts. Per [expr.static.cast]p2 and p11,
base-to-derived casts have undefined behavior if the object is not actually an
instance of the derived type.
Manuel Klimek [Wed, 13 Feb 2013 10:46:36 +0000 (10:46 +0000)]
An attempt to make the search algorithm easier to understand.
- clear ownership: the SpecificBumpPtrAllocator owns all StateNodes
- this allows us to simplify the memoization data structure into a
std::set (FIXME: figure out whether we want to use a hash based
data structure).
- introduces StateNode as recursive data structure, instead of using
Edge and the Seen-map combined to drill through the graph
- using a count to stabilize the penalty instead of relying on the
container
- pulled out a method to forward-apply states in the end
This leads to a ~40% runtime decrease on Nico's benchmark.
Main FiXME is that the parameter lists of some function get too long.
I'd vote for either pulling the Queue etc into the Formatter proper,
or creating an inner class just for the search algorithm.
Jordan Rose [Wed, 13 Feb 2013 03:11:06 +0000 (03:11 +0000)]
[analyzer] Use Clang's evaluation for global constants and default arguments.
Previously, we were handling only simple integer constants for globals and
the smattering of implicitly-valued expressions handled by Environment for
default arguments. Now, we can use any integer constant expression that
Clang can evaluate, in addition to everything we handled before.
Dmitri Gribenko [Tue, 12 Feb 2013 17:27:41 +0000 (17:27 +0000)]
Accept over-qualified constructor in MSVC emulation mode
MSVC accepts this:
class A {
A::A();
};
Clang accepts regular member functions with extra qualification as an MS
extension, but not constructors. This changes the parser to defer rejecting
qualified constructors so that the same Sema logic can apply to constructors as
regular member functions. This also improves the error message when MS
extensions are disabled (in my opinion). Before it was:
/Users/jason/Desktop/test.cpp:2:8: error: expected member name or ';' after declaration specifiers
A::A();
~~~~ ^
1 error generated.
After:
/Users/jason/Desktop/test.cpp:2:6: error: extra qualification on member 'A'
A::A();
~~~^
1 error generated.
Nico Weber [Tue, 12 Feb 2013 16:17:07 +0000 (16:17 +0000)]
Formatter: Correctly format stars in `sizeof(int**)` and similar places.
This redoes how '*' and '&' are classified as pointer / reference markers when
followed by ')', '>', or ','.
Previously, determineStarAmpUsage() marked a single '*' and '&' followed by
')', '>', or ',' as pointer or reference marker. Now, all '*'s and '&'s
preceding ')', '>', or ',' are marked as pointer / reference markers. Fixes
PR14884.
Since only the last '*' in 'int ***' was marked as pointer before (the rest
were unary operators, which don't reach spaceRequiredBetween()),
spaceRequiredBetween() now had to be thought about handing multiple '*'s in
sequence.
Nick Lewycky [Tue, 12 Feb 2013 08:08:54 +0000 (08:08 +0000)]
The meat of this patch is in BuildCXXMemberCalLExpr where we make it use
MarkMemberReferenced instead of marking functions referenced directly. An audit
of callers to MarkFunctionReferenced and DiagnoseUseOfDecl also caused a few
other changes:
* don't mark functions odr-used when considering them for an initialization
sequence. Do mark them referenced though.
* the function nominated by the cleanup attribute should be diagnosed.
* operator new/delete should be diagnosed when building a 'new' expression.
Richard Smith [Tue, 12 Feb 2013 05:48:23 +0000 (05:48 +0000)]
Fix a bug reduced from a crash when trying to use modules with libc++. We check
the linkage of functions and variables while merging declarations from modules,
and we don't necessarily have enough of the rest of the AST loaded at that
point to allow us to compute linkage, so serialize it instead.
Richard Smith [Tue, 12 Feb 2013 02:32:35 +0000 (02:32 +0000)]
Remove an assert which triggers when a decl context in a module hits the 'has
lexical storage but not visible storage' case in C++. It's unclear whether we
even need the special-case handling for C++, since it seems to be working
around our not serializing a lookup table for the TU in C. But in any case,
the assertion is incorrect.
Lang Hames [Mon, 11 Feb 2013 23:44:11 +0000 (23:44 +0000)]
When generating IR for default copy-constructors, copy-assignment operators,
move-constructors and move-assignment operators, use memcpy to copy adjacent
POD members.
Previously, classes with one or more Non-POD members would fall back on
element-wise copies for all members, including POD members. This often
generated a lot of IR. Without padding metadata, it wasn't often possible
for the LLVM optimizers to turn the element-wise copies into a memcpy.
This code hasn't yet received any serious tuning. I didn't see any serious
regressions on a self-hosted clang build, or any of the nightly tests, but
I think it's important to get this out in the wild to get more testing.
Insights, feedback and comments welcome.
Many thanks to David Blaikie, Richard Smith, and especially John McCall for
their help and feedback on this work.
Richard Smith [Mon, 11 Feb 2013 22:02:16 +0000 (22:02 +0000)]
Don't bother reconciling external visible decls against our current set of
declarations if we didn't have a lookup map when the external decls were added.
Douglas Gregor [Mon, 11 Feb 2013 18:16:18 +0000 (18:16 +0000)]
[Modules] Cope better with top-level declarations loaded after being declared in the current translation unit <rdar://problem/13189985>.
These two related tweaks to keep the information associated with a
given identifier correct when the identifier has been given some
top-level information (say, a top-level declaration) and more
information is then loaded from a module. The first ensures that an
identifier that was "interesting" before being loaded from an AST is
considered to be different from its on-disk counterpart. Otherwise, we
lose such changes when writing the current translation unit as a
module.
Second, teach the code that injects AST-loaded names into the
identifier chain for name lookup to keep the most recent declaration,
so that we don't end up confusing our declaration chains by having a
different declaration in there.
Nico Weber [Mon, 11 Feb 2013 15:32:15 +0000 (15:32 +0000)]
Formatter: Detect ObjC message expressions after 'in' in loop
Before:
for (id foo in[self getStuffFor : bla]) {
}
Now:
for (id foo in [self getStuffFor:bla]) {
}
"in" is treated as loop keyword if the line starts with "for", and as a
regular identifier else. To check for "in", its IdentifierInfo is handed
through a few layers.
Daniel Jasper [Mon, 11 Feb 2013 12:36:37 +0000 (12:36 +0000)]
Fix invalid formatting with spaces before trailing comments.
In google style, trailing comments are separated by two spaces. This
patch fixes the counting of these spaces and prevents clang-format from
creating a line with 81 columns.
Nico Weber [Sun, 10 Feb 2013 20:35:35 +0000 (20:35 +0000)]
Formatter: Initial support for ObjC dictionary literals.
Before:
@{
foo:
bar
}
;
Now:
@{ foo : bar };
parseBracedList() already does the right thing from an UnwrappedLineParser
perspective, so check for "@{" in all loops that process constructs that can
contain expressions and call parseBracedList() if found.
Bob Wilson [Sun, 10 Feb 2013 16:01:41 +0000 (16:01 +0000)]
Use -mno-implicit-float by default for kernel/kext code. <rdar://13177960>
Apple's kernel engineers have been expecting this behavior even though
we've never implemented it before, as far as I can tell. In recent months,
clang has gotten better at using vector instructions to optimize memcpy-like
operations, and that has exposed problems when vector/floating-point
instructions are used in kexts that don't support that. This behavior also
matches what Apple's GCC did for PowerPC targets.
Bob Wilson [Sun, 10 Feb 2013 15:25:44 +0000 (15:25 +0000)]
Recognize -mno-implicit-float option for x86 as well as ARM. <rdar://13180731>
For x86 targets, we've been using the -msoft-float option to control passing
the no-implicit-float option to cc1. Since the -mno-implicit-float option is
now accepted by the driver, this just makes it work for x86 the same as it
does for ARM targets.
objective-C: Fixes a bogus warning due to not setting
the "nonatomic" attribute in property redeclaration
in class extension. Also, improved on diagnostics in
this area while at it. // rdar://13156292
Jordan Rose [Sat, 9 Feb 2013 02:12:23 +0000 (02:12 +0000)]
Release notes: mention support for Unicode and UCNs in identifiers.
I'm using the name "Extended Identifiers" for the feature because that's
what GCC calls them. According to the standard, the new feature is
"universal character names are now allowed in identifiers", but the more
interesting "feature" is that identifiers can now contain Unicode characters,
however they are written.
Douglas Gregor [Sat, 9 Feb 2013 01:35:03 +0000 (01:35 +0000)]
Ensure that type definitions present in just-loaded modules are
visible.
The basic problem here is that a given translation unit can use
forward declarations to form pointers to a given type, say,
class X;
X *x;
and then import a module that includes a definition of X:
import XDef;
We will then fail when attempting to access a member of X, e.g.,
x->method()
because the AST reader did not know to look for a default of a class
named X within the new module.
This implementation is a bit of a C-centric hack, because the only
definitions that can have this property are enums, structs, unions,
Objective-C classes, and Objective-C protocols, and all of those are
either visible at the top-level or can't be defined later. Hence, we
can use the out-of-date-ness of the name and the identifier-update
mechanism to force the update.
In C++, we will not be so lucky, and will need a more advanced
solution, because the definitions could be in namespaces defined in
two different modules, e.g.,
// module 1
namespace N { struct X; }
// module 2
namespace N { struct X { /* ... */ }; }
One possible implementation here is for C++ to extend the information
associated with each identifier table to include the declaration IDs
of any definitions associated with that name, regardless of
context. We would have to eagerly load those definitions.
Jordan Rose [Sat, 9 Feb 2013 01:10:25 +0000 (01:10 +0000)]
Properly validate UCNs for C99 and C++03 (both more restrictive than C(++)11).
Add warnings under -Wc++11-compat, -Wc++98-compat, and -Wc99-compat when a
particular UCN is incompatible with a different standard, and -Wunicode when
a UCN refers to a surrogate character in C++03.
Anna Zaks [Sat, 9 Feb 2013 01:09:27 +0000 (01:09 +0000)]
[analyzer] Invalidation checker: move the "missing implementation" check
The missing definition check should be in the same category as the
missing ivar validation - in this case, the intent is to invalidate in
the given class, as described in the declaration, but the implementation
does not perform the invalidation. Whereas the MissingInvalidationMethod
checker checks the cases where the method intention is not to
invalidate. The second checker has potential to have a much higher false
positive rate.
projects / clang / log