Chris Lattner [Wed, 19 Nov 2008 06:51:40 +0000 (06:51 +0000)]
rewrite FormatDiagnostic to be less gross and a lot more efficient.
This also makes it illegal to have bare '%'s in diagnostics. If you
want a % in a diagnostic, use %%.
Chris Lattner [Wed, 19 Nov 2008 06:04:55 +0000 (06:04 +0000)]
implement a transparent optimization with the diagnostics stuff:
const char*'s are now not converted to std::strings when the diagnostic
is formed, we just hold onto their pointer and format as needed.
This commit makes DiagnosticClient::FormatDiagnostic even more of a
mess, I'll fix it in the next commit.
Chris Lattner [Wed, 19 Nov 2008 05:08:23 +0000 (05:08 +0000)]
Switch several more Sema Diag methods over. This simplifies the
__builtin_prefetch code to only emit one diagnostic per builtin_prefetch.
While this has nothing to do with the rest of the patch, the code seemed
like overkill when I was updating it.
Douglas Gregor [Wed, 19 Nov 2008 03:25:36 +0000 (03:25 +0000)]
Built-in equality and relational operators have return type "bool" in C++,
not "int".
Fix a typo in the promotion of enumeration types that was causing some
integral promotions to look like integral conversions (leading to
extra ambiguities in overload resolution).
Check for "acceptable" overloaded operators based on the types of the
arguments. This is a somewhat odd check that is specified by the
standard, but I can't see why it actually matters: the overload
candidates it suppresses don't seem like they would ever be picked as
the best candidates.
Ted Kremenek [Wed, 19 Nov 2008 01:46:21 +0000 (01:46 +0000)]
scan-build now recognizes when the user is directly invoking gcc for:
(a) different versions of gcc (e.g., a suffix '-4.2')
(b) different paths for gcc
Ted Kremenek [Wed, 19 Nov 2008 00:46:18 +0000 (00:46 +0000)]
- Move static function IsNonPragmaNonMacroLexer into Preprocessor.h.
- Add variants of IsNonPragmaNonMacroLexer to accept an IncludeMacroStack entry
(simplifies some uses).
- Use IsNonPragmaNonMacroLexer in Preprocessor::LookupFile.
Performance testing of -Eonly on Cocoa.h shows no performance regression because
of this patch.
Douglas Gregor [Tue, 18 Nov 2008 23:14:02 +0000 (23:14 +0000)]
Partial expansion of C++ operator overloading (for binary operators)
to support operators defined as member functions, e.g.,
struct X {
bool operator==(X&);
};
Overloading with non-member operators is supported, and the special
rules for the implicit object parameter (e.g., the ability for a
non-const *this to bind to an rvalue) are implemented.
This change also refactors and generalizes the code for adding
overload candidates for overloaded operator calls (C++ [over.match.expr]),
both to match the rules more exactly (name lookup of non-member
operators actually ignores member operators) and to make this routine
more reusable for the other overloaded operators.
Testing for the initialization of the implicit object parameter is
very light. More tests will come when we get support for calling
member functions directly (e.g., o.m(a1, a2)).
Douglas Gregor [Tue, 18 Nov 2008 15:03:34 +0000 (15:03 +0000)]
As threatened previously: consolidate name lookup and the creation of
DeclRefExprs and BlockDeclRefExprs into a single function
Sema::ActOnDeclarationNameExpr, eliminating a bunch of duplicate
lookup-name-and-check-the-result code.
Note that we still have the three parser entry points for identifiers,
operator-function-ids, and conversion-function-ids, since the parser
doesn't (and shouldn't) know about DeclarationNames. This is a Good
Thing (TM), and there will be more entrypoints coming (e.g., for C++
pseudo-destructor expressions).
Douglas Gregor [Tue, 18 Nov 2008 14:39:36 +0000 (14:39 +0000)]
Extend DeclarationName to support C++ overloaded operators, e.g.,
operator+, directly, using the same mechanism as all other special
names.
Removed the "special" identifiers for the overloaded operators from
the identifier table and IdentifierInfo data structure. IdentifierInfo
is back to representing only real identifiers.
Added a new Action, ActOnOperatorFunctionIdExpr, that builds an
expression from an parsed operator-function-id (e.g., "operator
+"). ActOnIdentifierExpr used to do this job, but
operator-function-ids are no longer represented by IdentifierInfo's.
Extended Declarator to store overloaded operator names.
Sema::GetNameForDeclarator now knows how to turn the operator
name into a DeclarationName for the overloaded operator.
Except for (perhaps) consolidating the functionality of
ActOnIdentifier, ActOnOperatorFunctionIdExpr, and
ActOnConversionFunctionExpr into a common routine that builds an
appropriate DeclRefExpr by looking up a DeclarationName, all of the
work on normalizing declaration names should be complete with this
commit.
Chris Lattner [Tue, 18 Nov 2008 07:48:38 +0000 (07:48 +0000)]
Change a couple of the Parser::Diag methods to return DiagnosticInfo
and let the clients push whatever they want into the DiagnosticInfo
instead of hard coding a few forms. Also switch various clients to
use Diag(Tok, ...) instead of Diag(Tok.getLocation(), ...) as the
canonical form to simplify the code a bit.
Chris Lattner [Tue, 18 Nov 2008 07:04:44 +0000 (07:04 +0000)]
This reworks some of the Diagnostic interfaces a bit to change how diagnostics
are formed. In particular, a diagnostic with all its strings and ranges is now
packaged up and sent to DiagnosticClients as a DiagnosticInfo instead of as a
ton of random stuff. This has the benefit of simplifying the interface, making
it more extensible, and allowing us to do more checking for things like access
past the end of the various arrays passed in.
In addition to introducing DiagnosticInfo, this also substantially changes how
Diagnostic::Report works. Instead of being passed in all of the info required
to issue a diagnostic, Report now takes only the required info (a location and
ID) and returns a fresh DiagnosticInfo *by value*. The caller is then free to
stuff strings and ranges into the DiagnosticInfo with the << operator. When
the dtor runs on the DiagnosticInfo object (which should happen at the end of
the statement), the diagnostic is actually emitted with all of the accumulated
information. This is a somewhat tricky dance, but it means that the
accumulated DiagnosticInfo is allowed to keep pointers to other expression
temporaries without those pointers getting invalidated.
This is just the minimal change to get this stuff working, but this will allow
us to eliminate the zillions of variant "Diag" methods scattered throughout
(e.g.) sema. For example, instead of calling:
Chris Lattner [Tue, 18 Nov 2008 04:56:44 +0000 (04:56 +0000)]
Change the diagnostics interface to take an array of pointers to
strings instead of array of strings. This reduces string copying
in some not-very-important cases, but paves the way for future
improvements.
Ted Kremenek [Tue, 18 Nov 2008 01:33:13 +0000 (01:33 +0000)]
- Add Lexer::isPragma() accessor for clients of Lexer that aren't friends.
- Add static method to test if the current lexer is a non-macro/non-pragma
lexer.
- Refactor some code in PPLexerChange to use this static method.
- No performance change.
Ted Kremenek [Tue, 18 Nov 2008 01:04:47 +0000 (01:04 +0000)]
Add hooks to use PTHLexer::Lex instead of Lexer::Lex when CurLexer is null.
Performance tests on Cocoa.h (using the regular Lexer) shows no performance
difference.
Ted Kremenek [Tue, 18 Nov 2008 00:43:07 +0000 (00:43 +0000)]
Added conditional guard 'if (CurLexer)' when using SetCommentRetentionState().
This is because the PTHLexer will not support this method. Performance testing
on preprocessing Cocoa.h shows that this results in a negligible performance
difference (less than 1%).
I tried making Lexer::SetCommentRetentionState() an out-of-line function (a
precursor to making it a virtual function in PreprocessorLexer) and noticed a 1%
decrease in speed (it is called in a hot part of the Preprocessor).
Ted Kremenek [Tue, 18 Nov 2008 00:34:22 +0000 (00:34 +0000)]
Change a bunch of uses of 'CurLexer->' to 'CurPPLexer->', which should be the
alias for the current PreprocessorLexer. No functionality change. Performance
testing shows this results in no performance degradation when preprocessing
Cocoa.h.
Ted Kremenek [Tue, 18 Nov 2008 00:12:49 +0000 (00:12 +0000)]
- Add 'CurPPLexer' to Preprocessor to keep track of the current
PreprocessorLexer, which will either be a 'Lexer' or 'PTHLexer'.
- Added stub field 'CurPTHLexer' to keep track of the current PTHLexer.
- Modified IncludeStackInfo to track both the current PTHLexer and
current PreprocessorLexer.
Douglas Gregor [Mon, 17 Nov 2008 22:58:34 +0000 (22:58 +0000)]
Eliminate all of the placeholder identifiers used for constructors,
destructors, and conversion functions. The placeholders were used to
work around the fact that the parser and some of Sema really wanted
declarators to have simple identifiers; now, the code that deals with
declarators will use DeclarationNames.
Ted Kremenek [Mon, 17 Nov 2008 22:55:12 +0000 (22:55 +0000)]
Fold assertion into second valid else branch. This removes a compiler warning
where the control reaches the end of a non-void function and also allows the
compiler to generate better code. When this assertion is false we can easily
add more else cases.
Douglas Gregor [Mon, 17 Nov 2008 20:34:05 +0000 (20:34 +0000)]
Updated IdentifierResolver to deal with DeclarationNames. The names of
C++ constructors, destructors, and conversion functions now have a
FETokenInfo field that IdentifierResolver can access, so that these
special names are handled just like ordinary identifiers. A few other
Sema routines now use DeclarationNames instead of IdentifierInfo*'s.
To validate this design, this code also implements parsing and
semantic analysis for id-expressions that name conversion functions,
e.g.,
return operator bool();
The new parser action ActOnConversionFunctionExpr takes the result of
parsing "operator type-id" and turning it into an expression, using
the IdentifierResolver with the DeclarationName of the conversion
function. ActOnDeclarator pushes those conversion function names into
scope so that the IdentifierResolver can find them, of course.
Chris Lattner [Mon, 17 Nov 2008 19:51:54 +0000 (19:51 +0000)]
Implement rdar://6319320: give a good diagnostic for cases where people
are trying to use the old GCC "casts as lvalue" extension. We don't and
will hopefully never support this.
Douglas Gregor [Mon, 17 Nov 2008 14:58:09 +0000 (14:58 +0000)]
Introduction the DeclarationName class, as a single, general method of
representing the names of declarations in the C family of
languages. DeclarationName is used in NamedDecl to store the name of
the declaration (naturally), and ObjCMethodDecl is now a NamedDecl.
projects / clang / log