Chris Lattner [Sat, 19 Feb 2011 22:28:58 +0000 (22:28 +0000)]
implement a tiny amount of codegen support for gnu array range
designators: allowing codegen when the element initializer is a
constant or something else without a side effect. This unblocks
enough to let process.c in the linux kernel build, PR9257.
Douglas Gregor [Sat, 19 Feb 2011 21:32:49 +0000 (21:32 +0000)]
Handle the resolution of a reference to a function template (which
includes explicitly-specified template arguments) to a function
template specialization in cases where no deduction is performed or
deduction fails. Patch by Faisal Vali, fixes PR7505!
Douglas Gregor [Sat, 19 Feb 2011 19:24:40 +0000 (19:24 +0000)]
The member classes of a current instantiation aren't necessarily a
current instantiation, even though we have a RecordDecl describing
them. Fixes PR9255.
Amusingly, I've had this patch sitting around for a month or two
because it was "obviously" wrong, but hadn't gotten around to writing
a test case to submit the fix :)
Douglas Gregor [Sat, 19 Feb 2011 19:14:36 +0000 (19:14 +0000)]
Teach the virtual-functions-without-virtual-destructor warning to only
warn about polymorphic classes (which have virtual functions) rather
than dynamic classes (which are polymorphic or have virtual bases).
Douglas Gregor [Sat, 19 Feb 2011 18:51:44 +0000 (18:51 +0000)]
Revert all of my commits that devirtualized the Decl hierarchy, which
lead to a serious slowdown (4%) on parsing of Cocoa.h. This memory
optimization should be revisited later, when we have time to look at
the generated code.
John McCall [Sat, 19 Feb 2011 03:13:26 +0000 (03:13 +0000)]
Fix a -Wuninitialized warning; it's actually a false positive,
but it's not reasonable for the diagnostic to figure that out.
Pointed out by Benjamin Kramer.
John McCall [Sat, 19 Feb 2011 02:53:41 +0000 (02:53 +0000)]
Warn about code that uses variables and functions with internal linkage
without defining them. This should be an error, but I'm paranoid about
"uses" that end up not actually requiring a definition. I'll revisit later.
Also, teach IR generation to not set internal linkage on variable
declarations, just for safety's sake. Doing so produces an invalid module
if the variable is not ultimately defined.
Also, fix several places in the test suite where we were using internal
functions without definitions.
Ted Kremenek [Sat, 19 Feb 2011 01:59:33 +0000 (01:59 +0000)]
Add 'StoreRef' smart pointer to allow more fine-grain memory lifetime control of Store objects.
This yields a minor memory reduction (for larger functions) on Sqlite at the cost of slightly
higher memory usage on some functions because of the increased size of GRState (which can be optimized).
I expect the real memory savings from this enhancement will come when we aggressively
canabilize more of the ExplodedGraph.
Douglas Gregor [Fri, 18 Feb 2011 23:30:37 +0000 (23:30 +0000)]
When code-completing a case statement for a switch on a value of
enumeration type, prioritize the enumeration constants and don't
provide completions for any other expressions. Fixes <rdar://problem/7283668>.
Douglas Gregor [Fri, 18 Feb 2011 22:29:55 +0000 (22:29 +0000)]
Selector::getIdentifierInfoForSlot() can return NULL values, a fact
that was ignored in a few places (most notably, code
completion). Introduce Selector::getNameForSlot() for the common case
where we only care about the name. Audit all uses of
getIdentifierInfoForSlot(), switching many over to getNameForSlot(),
fixing a few crashers.
Fixed <rdar://problem/8939352>, a code-completion crasher.
NAKAMURA Takumi [Fri, 18 Feb 2011 08:44:38 +0000 (08:44 +0000)]
[Heikki Kultala] This patch contains the ABI changes for the TCE target.
TCE target has some too strict alignment rules (that the HW really does not require, but which caused problems elsewhere) for data types and an ABI change was decided.
Douglas Gregor [Fri, 18 Feb 2011 03:01:41 +0000 (03:01 +0000)]
Implement C++0x [expr.static.cast]p9, which permits explicitly casting
a scoped enumeration type to an integral or floating type,
properly. There was an over-eager assertion, and it was missing the
floating-point case.
Douglas Gregor [Fri, 18 Feb 2011 02:44:58 +0000 (02:44 +0000)]
When building a qualified reference to a member of an anonymous struct
or union, place the qualifier on the outermost member reference
expression, which actually contains the entity name.
Move CompilerInstance::LLVMContext and LLVMContext ownership to CodeGenAction
This removes the final dependency edge from any lib outside of CodeGen
to core. As a result we can, and do, trim the dependency on core
from libclang, PrintFunctionNames, the unit tests and c-index-test.
While at it, review and trim other unneeded dependencies.
Douglas Gregor [Fri, 18 Feb 2011 02:12:44 +0000 (02:12 +0000)]
When we're creating an expression for an integral template argument of
enumeration type, we were generating an integer literal implicitly
casted to the appropriate enumeration type. However, later checks on
that expression would strip the implicit cast.
This commit tweaks the lame hack, by creating an explicit cast instead
of an implicit cast. The right answer is to introduce a
SubstNonTypeTemplateParmExpr expression that acts like the substituted
result. I'll investigate that soon.
Chris Lattner [Fri, 18 Feb 2011 02:08:43 +0000 (02:08 +0000)]
implement basic support for __label__. I wouldn't be shocked if there are
bugs from other clients that don't expect to see a LabelDecl in a DeclStmt,
but if so they should be easy to fix.
Chris Lattner [Fri, 18 Feb 2011 01:27:55 +0000 (01:27 +0000)]
Switch labels over to using normal name lookup, instead of their
own weird little DenseMap. Hey look, we now emit unused label
warnings deterministically, amazing.
[analyzer] Use the new registration mechanism for the debugging info "checks".
The relative checker package is 'debug':
'-dump-live-variables' is replaced by '-analyzer-checker=debug.DumpLiveVars'
'-cfg-view' is replaced by '-analyzer-checker=debug.ViewCFG'
'-cfg-dump' is replaced by '-analyzer-checker=debug.DumpCFG'
[analyzer]
-Introduce CheckerV2, a set of templates for convenient declaration & registration of checkers.
Currently useful just for checkers working on the AST not the path-sensitive ones.
-Enhance CheckerManager to actually collect the checkers and turn it into the entry point for
running the checkers.
-Use the new mechanism for the LLVMConventionsChecker.
Chris Lattner [Thu, 17 Feb 2011 20:34:02 +0000 (20:34 +0000)]
Step #2/N of __label__ support: keep pushing LabelDecl forward,
making them be template instantiated in a more normal way and
make them handle attributes like other decls.
This fixes the used/unused label handling stuff, making it use
the same infrastructure as other decls.
Douglas Gregor [Thu, 17 Feb 2011 19:04:40 +0000 (19:04 +0000)]
Eliminate the internal command-line option for viewing inheritance in C++ classes, since it's only really worked for the trivial cases anyway due to lame pseudo-parsing of the class name. The viewInheritance() function is still available for use in the debugger, where this is far more useful
When building StaticAnalyzer/Frontend add -I "<Checkers build dir>" to allow Checkers.inc to be
included without '..', thus being compatible with build systems of *BSDs.
Douglas Gregor [Thu, 17 Feb 2011 18:14:32 +0000 (18:14 +0000)]
Remove the last virtual member function from the Decl hierarchy,
reducing the size of all declarations by one pointer. For a 64-bit
Clang parsing Cocoa.h, this saves ~630k of memory (about 3.5% of
ASTContext's memory usage for this header).
Chandler Carruth [Thu, 17 Feb 2011 11:05:49 +0000 (11:05 +0000)]
Implement a sub-group of -Wconversion: -Wliteral-conversion. This
specifically targets literals which are implicitly converted, a those
are more often unintended and trivial to fix. This can be especially
helpful for diagnosing what makes 'const int x = 1e6' not an ICE.
Original patch authored by Jim Meehan with contributions from other
Googlers and a few cleanups from myself.
John McCall [Thu, 17 Feb 2011 10:25:35 +0000 (10:25 +0000)]
Change the representation of GNU ?: expressions to use a different expression
class and to bind the shared value using OpaqueValueExpr. This fixes an
unnoticed problem with deserialization of these expressions where the
deserialized form would lose the vital pointer-equality trait; or rather,
it fixes it because this patch also does the right thing for deserializing
OVEs.
Change OVEs to not be a "temporary object" in the sense that copy elision is
permitted.
This new representation is not totally unawkward to work with, but I think
that's really part and parcel with the semantics we're modelling here. In
particular, it's much easier to fix things like the copy elision bug and to
make the CFG look right.
I've tried to update the analyzer to deal with this in at least some
obvious cases, and I think we get a much better CFG out, but the printing
of OpaqueValueExprs probably needs some work.
projects / clang / log