This fixes the case where the wrong symbol is emitted leading to linking errors when you reference a class before defining it (GNU runtime).
Patch by David Chisnall.
Fix a horrible CFG bug reported in <rdar://problem/7027684>. The wrong successor
block would get hooked up in some cases when processing empty compound
statements.
Replace guarded calls in RegionStoreManager to
ValueManager::getRegionValueSymbolVal() with unguarded calls to
ValueManager::getRegionValueSymbolValOrUnknown(). This changes centralizes the
decision of what values to symbolicate in SymbolManager rather than having it
scatter in RegionStoreManager.
StoreManagers: Use 'hasGlobalsStorage()' and 'hasParametersStorage()' instead of
directly consulting if a VarDecl is an implicit or actual parameter, a global,
etc.
Add a separate MemSpaceRegion for function/method arguments passed on the stack.
This will simplify the logic of StoreManagers that want to specially reason
about the values of parameters.
Douglas Gregor [Thu, 2 Jul 2009 17:08:52 +0000 (17:08 +0000)]
Add support for retrieving the Doxygen comment associated with a given
declaration in the AST.
The new ASTContext::getCommentForDecl function searches for a comment
that is attached to the given declaration, and returns that comment,
which may be composed of several comment blocks.
Comments are always available in an AST. However, to avoid harming
performance, we don't actually parse the comments. Rather, we keep the
source ranges of all of the comments within a large, sorted vector,
then lazily extract comments via a binary search in that vector only
when needed (which never occurs in a "normal" compile).
Comments are written to a precompiled header/AST file as a blob of
source ranges. That blob is only lazily loaded when one requests a
comment for a declaration (this never occurs in a "normal" compile).
The indexer testbed now supports comment extraction. When the
-point-at location points to a declaration with a Doxygen-style
comment, the indexer testbed prints the associated comment
block(s). See test/Index/comments.c for an example.
Some notes:
- We don't actually attempt to parse the comment blocks themselves,
beyond identifying them as Doxygen comment blocks to associate them
with a declaration.
- We won't find comment blocks that aren't adjacent to the
declaration, because we start our search based on the location of
the declaration.
- We don't go through the necessary hops to find, for example,
whether some redeclaration of a declaration has comments when our
current declaration does not. Similarly, we don't attempt to
associate a \param Foo marker in a function body comment with the
parameter named Foo (although that is certainly possible).
- Verification of my "no performance impact" claims is still "to be
done".
Temporarily disable RegionStore for stack-addr-ps.c, as a new test case reveals
a case where RegionStore doesn't create symbolic values for the fields of
structs that are passed-by-value.
Add a FIXME to RegionStore, do some minor code cleanup, and get RegionStore to
pass misc-ps.m. Currently RegionStore/BasicStore don't do any special reasoning
about clang-style vectors, so we should return UnknownVal (in all cases) when
accessing their values via an array.
Douglas Gregor [Wed, 1 Jul 2009 22:01:06 +0000 (22:01 +0000)]
Keep track of more information within the template instantiation stack, e.g.,
by distinguishing between substitution that occurs for template
argument deduction vs. explicitly-specifiad template arguments. This
is used both to improve diagnostics and to make sure we only provide
SFINAE in those cases where SFINAE should apply.
In addition, deal with the sticky issue where SFINAE only considers
substitution of template arguments into the *type* of a function
template; we need to issue hard errors beyond this point, as
test/SemaTemplate/operator-template.cpp illustrates.
Daniel Dunbar [Wed, 1 Jul 2009 20:03:04 +0000 (20:03 +0000)]
Driver: Move Compilation::Execute to Driver::ExecuteCompilation.
- The Compilation is just a helper class, it shouldn't have that amount of
logic in it.
Daniel Dunbar [Wed, 1 Jul 2009 19:14:39 +0000 (19:14 +0000)]
Driver: Improve diagnostics for failed commands.
- Not all tools give good error messages, and sometimes the tool can fail w/o
any error (for example, when signalled).
- We suppress this message when the failing command is the compiler and it
failed normally (exit code == 1), under the assumption that it gave a good
diagnostic.
For example, for a linker failure we now get:
--
ddunbar@lordcrumb:tmp$ clang a.c b.c
ld: duplicate symbol _x in /var/folders/cl/clrOX6SaG+moCeRKEI4PtU+++TI/-Tmp-/cc-bXYITq.o and /var/folders/cl/clrOX6SaG+moCeRKEI4PtU+++TI/-Tmp-/cc-6uK4jD.o
clang: error: linker command failed with exit code 1 (use -v to see invocation)
--
For a compiler crash we get:
--
ddunbar@lordcrumb:tmp$ clang t.i
Assertion failed: (CGT.getTargetData().getTypeAllocSizeInBits(STy) == RL.getSize()), function layoutStructFields, file CodeGenTypes.cpp, line 573.
0 clang-cc 0x0000000100f1f1f1 PrintStackTrace(void*) + 38
... stack trace and virtual stack trace follow ...
clang: error: compiler command failed due to signal 6 (use -v to see invocation)
--
But for a regular compilation failure we get the usual:
--
ddunbar@lordcrumb:tmp$ clang c.c
c.c:1:6: error: invalid token after top level declarator
int x
^
1 diagnostic generated.
--
- No test case, not amenable to non-executable testing. :/
Daniel Dunbar [Wed, 1 Jul 2009 19:02:28 +0000 (19:02 +0000)]
Driver: Add Source argument for Commands to hold the Action which caused a
Command to be generated, to support more advanced diagnostics.
- No functionality change.
Eli Friedman [Wed, 1 Jul 2009 03:36:11 +0000 (03:36 +0000)]
Fix for PR4192: fix the definition of int64_t on x86_64 Linux.
Note that I'm guessing that *BSD and Solaris do the same thing as Linux
here, but it's quite possible I'm wrong; if the following testcase
gives an error on x86-64 with gcc for any of those operating systems, please
tell me:
Douglas Gregor [Wed, 1 Jul 2009 00:28:38 +0000 (00:28 +0000)]
Cope with explicitly-specified function template arguments when there
are fewer template arguments than there are template parameters for
that function.
Douglas Gregor [Tue, 30 Jun 2009 23:57:56 +0000 (23:57 +0000)]
When explicit template arguments are provided for a function call,
substitute those template arguments into the function parameter types
prior to template argument deduction. There's still a bit of work to
do to make this work properly when only some of the template arguments
are specified.
Douglas Gregor [Tue, 30 Jun 2009 22:34:41 +0000 (22:34 +0000)]
Preliminary parsing and ASTs for template-ids that refer to function
templates, such as make<int&>. These template-ids are only barely
functional for function calls; much more to come.
Ted Kremenek [Tue, 30 Jun 2009 22:31:23 +0000 (22:31 +0000)]
Update old CastRegion logic to not assume that ElementRegion's super region is a
TypedRegion. While we plan on removing this code at some point, it serves as a
good reference implementation for use with BasicStore until we are sure the new
CastRegion logic (in RegionStore.cpp) is correct.
Douglas Gregor [Tue, 30 Jun 2009 17:20:14 +0000 (17:20 +0000)]
When recursively instantiating function templates, keep track of the
instantiation stack so that we provide a full instantiation
backtrace. Previously, we performed all of the instantiations implied
by the recursion, but each looked like a "top-level" instantiation.
The included test case tests the previous fix for the instantiation of
DeclRefExprs. Note that the "instantiated from" diagnostics still
don't tell us which template arguments we're instantiating with.
Torok Edwin [Tue, 30 Jun 2009 16:30:17 +0000 (16:30 +0000)]
Use templates instead of creating a new class for each OS/Target combination.
This unifies all the targets supported by an OS into a template.
It also cleans up the differences between the darwin targets.
Also __LP64__ wasn't needed for *BSD, since x86-64 target defines it anyway.
Douglas Gregor [Tue, 30 Jun 2009 15:47:41 +0000 (15:47 +0000)]
Refactor ActOnDeclarationNameExpr into a "parsing action" part and a
"semantic analysis" part. Use the "semantic analysis" part when
performing template instantiation on a DeclRefExpr, rather than an ad
hoc list of rules to construct DeclRefExprs from the instantiation.
A test case for this change will come in with a large commit, which
illustrates what I was actually trying to work on.
Zhongxing Xu [Tue, 30 Jun 2009 13:00:53 +0000 (13:00 +0000)]
Instead of r74522, use another approach to fix xfail_regionstore_wine_crash.c.
Mark the super region of the binding of block level expr in the Environment
as live.
Zhongxing Xu [Tue, 30 Jun 2009 12:11:58 +0000 (12:11 +0000)]
Block level expr should be visited. Otherwise variables in init expr of
DeclStmt would be dead before the DeclStmt.
For example:
int x = 0;
int y = x;
'x' would be dead before 'int y = x'.
Remove ASTContext parameter from DeclContext's methods. This change cascaded down to other Decl's methods and changes to call sites started "escalating".
Timings using pre-tokenized "cocoa.h" showed only a ~1% increase in time run between and after this commit.
Remove the ASTContext parameter from the attribute-related methods of Decl.
The implementations of these methods can Use Decl::getASTContext() to get the ASTContext.
This commit touches a lot of files since call sites for these methods are everywhere.
I used pre-tokenized "carbon.h" and "cocoa.h" headers to do some timings, and there was no real time difference between before the commit and after it.
Chris Lattner [Tue, 30 Jun 2009 01:26:17 +0000 (01:26 +0000)]
Key decisions about 'bool' vs '_Bool' to be based on a new flag in langoptions.
This is simple enough, but then I thought it would be nice to make PrintingPolicy
get a LangOptions so that various things can key off "bool" and "C++" independently.
This spiraled out of control. There are many fixme's, but I think things are slightly
better than they were before.
One thing that can be improved: CFG should probably have an ASTContext pointer in it,
which would simplify its clients.
Douglas Gregor [Mon, 29 Jun 2009 22:39:32 +0000 (22:39 +0000)]
Improve code generation for function template specializations:
- Track implicit instantiations vs. the not-yet-supported explicit
specializations
- Give implicit instantiations of function templates (and member
functions of class templates) linkonce_odr linkage.
- Improve name mangling for function template specializations,
including the template arguments of the instantiation and the return
type of the function.
Note that our name-mangling is improved, but not correct: we still
don't mangle substitutions, although the manglings we produce can be
demangled.
Douglas Gregor [Mon, 29 Jun 2009 20:59:39 +0000 (20:59 +0000)]
Keep track of function template specializations, to eliminate
redundant, implicit instantiations of function templates and provide a
place where we can hang function template specializations.
-Keep a reference to the ASTContext inside the TranslationUnitDecl.
-Introduce Decl::getASTContext() which returns the reference from the TranslationUnitDecl that it is contained in.
The general idea is that Decls can point to their own ASTContext so that it is no longer required to "manually" keep track and make sure that you pass the correct ASTContext to Decls' methods, e.g. methods like Decl::getAttrs should eventually not require a ASTContext parameter.
projects / clang / log