Submitted by:
Reviewed by:
A bunch of "small" changes...
- Fixed a bug in ConvertFloatingRankToComplexType() that rendered it useless.
ASTContext was being copied by default (as the result of declaring the argument
incorrectly). Chris gave me the magic potion to disallow this in ASTContext.
- Removed a tab:-)
- Added some much needed comments to the float/complex promotion madness.
- Improved some aesthetics (based on code review w/Chris).
llvm-svn: 39414
Submitted by:
Reviewed by:
Refactored code that deals with float/complex conversions (the previous commit).
- Removed 6 predicates that identify the built-in float/complex types.
- Added two helper functions GetFloatingRank() & ConvertFloatingRankToComplexType().
At present, these are static functions in SemaExpr.cpp. Conceptually, they would be nice
to move to Type, however there are layering problems with that (i.e. no ASTContext).
Another possibility is to move them to ASTContext?
- Simplified the logic in UsualArithmeticConversions() considerably.
llvm-svn: 39413
Submitted by:
Reviewed by:
-Lot's of detail work in UsualArithmeticConversion(). Needed to expand
the code for dealing with floating types. This code still has a couple
of FIXME's and could be refactored a bit.
-Added a bunch of Type predicates to simplify the float conversions.
Used the opportunity to convert a bunch of predicates to dyn_cast (it's cleaner
when just dealing with Builtin types).
llvm-svn: 39412
Submitted by:
Reviewed by:
Lot's of changes related to type checking binary expressions.
- Changed the name/proto of ImplicitConversion. It is now named UsualUnaryConversion - it takes a
type and returns a type. This allowed me to remove the explicit node creation for array/function->pointer conversions.
- Added function UsualArithmeticConversions().
- Changed all the "Check" functions for binary ops. They use the new "Usual" functions above.
- new predicates on Type, isSignedIntegerType()/isUnsignedIntegerType().
- moved getDecl() out of the Sema class. It is now a static helper function in SemaExpr.cpp. It was also renamed getPrimaryDeclaration().
- Added CheckArithmeticOperand() for consistency with the other unary "Check" functions.
Should finish up the binary expressions tomorrow...with a small number of "FIXME's"
llvm-svn: 39411
Submitted by:
Reviewed by:
Start hacking on binary ops...
Important bug fix...was ignoring the return value from ImplicitConversion.
This didn't bother the type checking logic, however the AST was malformed.
llvm-svn: 39410
Submitted by:
Reviewed by:
Continue working on type checking for unary operators. Added:
- Two 3 private functions to Sema: CheckAddressOfOperand(), CheckIndirectionOperand(),
and getDecl().
- Added Expr::isLvalue() - it was needed for CheckAddressOfOperand(). It will
also be needed for ++/-- and the assignment operators.
- Added a couple diagnostics for invalid lvalues (for & of).
llvm-svn: 39408
implements test/Preprocessor/expr_usual_conversions.c, which produces this
output:
expr_usual_conversions.c:5:10: warning: left side of operator converted from negative value to unsigned: -42 to 18446744073709551574
#if (-42 + 0U) / -2
^
expr_usual_conversions.c:5:16: warning: right side of operator converted from negative value to unsigned: -2 to 18446744073709551614
#if (-42 + 0U) / -2
^
llvm-svn: 39406
like this:
t3.c:5:10: error: 'vers2.h' file not found
#include xstr(INCFILE(2).h)
^
instead of:
t3.c:5:10: error: file not found
#include xstr(INCFILE(2).h)
^
which is useful if the #include name is generated through macro expansion.
llvm-svn: 39398
Submitted by:
Reviewed by:
-Changed the name of TypeRef to QualType. Many diffs.
-Changed the QualType constructor to require Quals be passed. This makes the code a bit
more verbose, however will make the code easier to work on. Given the complexity
of types, this should help spot bogosities.
-Changed the Expr constructor to require a QualType. Same motivation.
llvm-svn: 39395
Submitted by:
Reviewed by:
Fix "FIXME: does this lose qualifiers from the typedef??" in ASTContext::getTypedefType().
This change was fairly pervasive...nevertheless, here are the highlights:
- Change the type of Type::CanonicalType to TypeRef (was "Type *").
- Change the implementation of TypeRef::getCanonicalType() to work for typedefs.
- Make the implementation of Type::getCanonicalType private (only TypeRef should access). This
will force clients to use TypeRef::getCanonicalType (the correct version of the function). Since
TypeRef overloads "->", it was very easy to fall into this bug...
- Changed many references of "Type *" to "TypeRef"...when the raw type pointer is required, use t.getTypePtr().
- Changed all the *Type classes to take a TypeRef.
- Made the Type constructor protected (cleanup).
- Removed function Expr::getType().
- Convert functions in SemaExpr to use the above support. This fixed the "const" bug I was originally investigating.
I will follow this check-in up with a rename of TypeRef->QualType. I will also make sure the constructor does not default to 0 (which can lead to broken code...).
llvm-svn: 39394
constant expressions. This allows us to emit this diagnostic:
t.c:5:5: warning: integer constant is so large that it is unsigned
#if 12345678901234567890
^
And makes constant evaluation fully correct, but we do not yet detect and
warn about integer overflow.
This patch requires cvs up'ing the main llvm tree to get the APSInt class,
but no libraries need to be rebuilt there.
llvm-svn: 39388
sized to the target's intmax_t. This also allows us to emit diagnostic like:
t.c:11:5: warning: integer constant is too large for its type
#if 18446744073709551616 // 2^64
^
t.c:15:5: warning: integer constant is too large for its type
#if 18446744073709551617 // 2^64-1
^
llvm-svn: 39385
Submitted by:
Reviewed by:
- Typechecking for ++/-- not operating on canonical type (fixed).
- Many changes to Decl.h...
-- Changed ObjectDecl->ValueDecl.
-- ValueDecl doesn't define a storage class. The storage class is now
defined by VarDecl/FunctionDecl. EnumConstant decls don't need a storage class.
-- FieldDecl now inherits directly from Decl (it is not a value).
- Changed SemaExpr to use the new Decl.h. Also added a couple diagnostics for
bogus storage class usage. For example:
z.c:6:10: error: illegal storage class on function
auto int func();
^
z.c:7:14: error: illegal storage class on file-scoped variable
register int b;
^
llvm-svn: 39379
Submitted by:
Reviewed by:
Incorporate code review feedback from Chris...
- minor restructure of ParseMemberReferenceExpr logic. The last statement
is now the instantiation of the AST node (which I like a lot and will conform
to throughout Sema).
- declare StmtClassNameTable const.
- reword an error diagnostic.
- install the correct type for ParseSizeOfAlignOfTypeExpr. Added hook in
ASTContext. For now, simply return Context.UnsignedLongTy. Added a FIXME
to revisit (i.e. compute using TargetInfo).
llvm-svn: 39374
Submitted by:
Reviewed by:
Add three new classes to Decl.h: BlockVarDecl, FileVarDecl, and ParmVarDecl.
Made the constructor to VarDecl protected, to indicate it is "abstract".
Reorganized the Decl::Kind enum to allow us to add ObjectDecls/TypeDecls with
minimal breakage. In the process, realized the isa support for ObjectDecl was already broken (so
the reorg paid for itself already:-) The range check should also be more efficient...
llvm-svn: 39373
Submitted by:
Reviewed by:
- ParseMemberReferenceExpr wasn't operating on the canonical type. From
now on, I will make sure the prologue to each Parse/Check function has
both the qualified type and the canonical type.
- More refinements to ParseDeclarator. It was allowing variable declarations
to incomplete types (e.g. void, struct foo, where foo wasn't defined).
llvm-svn: 39371
Submitted by:
Reviewed by:
Incorporate feedback from Chris (on the last check-in).
- added a shared hook for pre/post ++/-- CheckIncrementDecrementOperand().
- handle invalid arithmetic on incomplete types (void *, struct foo *, where the
body isn't in scope). Added a diagnostic for this case.
- added some comments and C99 annotations.
- Sema.h now includes Expr.h. I'd prefer not to, however it doesn't break
any layering.
llvm-svn: 39370
Submitted by:
Reviewed by:
Implement ++/-- typechecking for pre/post unary expressions. This includes:
- added isLvalue, isModifiableLvalue (on TypeRef, Type, and RecordType).
- added isRealType, isRealFloatingType, isComplexType.
- hacked Diag to take a TypeRef (I was sick of writing the 2 line "setup":-)
In addition, this will likely lead to less bugs...I already had written code
that was doing a getAsString on "Type" (which is wrong...since it doesn't include
any qualifiers).
- Changed UnaryOperator to take a TypeRef...pass it the right stuff.
- Removed redundant ternary expressions in several predicates.
- A couple diagnostics.
llvm-svn: 39369
Submitted by:
Reviewed by:
Finish up Sema::ParseArraySubscriptExpr. This involved:
- adding a couple predicates to Type.h (isObjectType, isDerivedType).
- added a diagnostic for subscripting non-object types (e.g. void (*)()).
- pass the correct result type...a minor detail:-)
- added some spec references to Type.h
llvm-svn: 39368
Submitted by:
Reviewed by:
Fix a couple bugs in Sema::GetTypeForDeclarator(). Need to disallow:
- void arrayOfFunctions[3]()
- void arrayOfVoids[3]
Need to talk to Chris about the return value...
llvm-svn: 39365
Submitted by:
Reviewed by:
Finish up Sema::ParseMemberReferenceExpr. This involved:
- added a getMember() function to RecordDecl.
- added stronger typing for "Members" (from Decl->FieldDecl).
- added a dignostic for members not found.
- changed MemberExpr to install the correct TypeRef.
- In general, simplified and cleaned up the routing.
llvm-svn: 39364
Submitted by:
Reviewed by:
Type Checking...round 2. This checkin "breaks" parsing carbon.h. I imagine
that this will be true for the next week or so. Nevertheless, this round of
changes includes the following:
- Hacked various Expr classes to pass the appropriate TypeRef. Still have
a few more classes to touch.
- Implement type checking for ParseArraySubscriptExpr and ParseMemberReferenceExpr.
- Added a debug hook to derive the class name for Stmt/Expr nodes. Currently a
linear search...could easily optimize if important.
- Changed the name of TaggedType->TagType. Now we have TagType and TagDecl (which
are easier to remember).
- Fixed a bug in StringLiteral conversion I did a couple weeks ago. hadError was
not initialized (oops).
- changed Sema::Diag to return true. This streamlines the type checking code
considerably.
- Added many diagnositics.
This should be it!
llvm-svn: 39361
Submitted by:
Reviewed by:
Implement type checking. First round of changes are:
- Added predicates to Type.
- Added predicates to BinExpr.
- Added Check hooks that model the categories for Binary ops.
- Added TypeRef to Expr. Will lazily eval subclasses...
- Misc bug fixes/cleanups.
llvm-svn: 39360
Submitted by:
Reviewed by:
carbon.h looking good! Only 1 warning left...no parse errors!
This fixes 3 bugs...
- A couple tricky bugs with type canonicalization. Nested typedef's weren't being
handled properly. For example, the following didn't work:
typdef int __darwin_pid_t;
typedef __darwin_pid_t pid_t;
int getpgid(pid_t);
int getpgid(int);
- The storage class wasn't being preserved. As a result, Sema was complaining
about the following:
extern char *foo;
char *foo;
- various built-ins weren't registered...resulting in spurious warnings.
llvm-svn: 39357
Submitted by:
Reviewed by:
Added size expression to ArrayType. This was needed to implement
Type::isIncompleteType(). At the moment, there is no support for
determining if we have a constant expression (which won't be too
difficult now that we have support for literal parsing/ast's).
Nevertheless, the parser will allow "struct foo { int a[b]; }"
(which is illegal). There is still significant work to fully analyze
array types. The good news is "carbon.h" goes from 288 bogus errors
down to 26!
llvm-svn: 39355
Submitted by:
Reviewed by:
Move string literal parsing from Sema=>LiteralSupport. This consolidates
all the quirky parsing code within the Lexer subsystem (yeah!). This
simplifies Sema and (more importantly) allows future parsers
(i.e. subclasses of Action) to benefit from this code.
llvm-svn: 39354
Submitted by:
Reviewed by:
-Converted the preprocessor to use NumericLiteralParser.
-Several minor changes to LiteralSupport interface/implementation.
-Added an error diagnostic for floating point usage in pp expr's.
llvm-svn: 39352
Submitted by:
Reviewed by:
Moved numeric literal support from SemaExpr.cpp to LiteralSupport.[h,cpp]
in Lex. This will allow it to be used by both Sema and Preprocessor (and
should be the last major refactoring of this sub-system).. Over
time, it will be reused by anyone implementing an actions module (i.e.
any subclass of llvm::clang::Action. Minor changes to IntegerLiteral in Expr.h.
More to come...
llvm-svn: 39351
Submitted by:
Reviewed by:
Refactored ParseNumericConstant using SkipDigits and friends. This
cleaned up the code and removed many "breaks". In the process, I found
several bugs that needed to be fixed (proof that this refactoring had
real value, not simply aesthetics). Also simplified error handling logic
and removed an unused error condition from Diagnostics.def.
llvm-svn: 39350
Submitted by:
Reviewed by:
More code to parse numeric constants. This checkin includes:
- Feedback from Chris.
- Support for parsing floating point constants.
- Moved the code to "Sema". Changed API in Action.
- More/better error diagnostics.
At this point, the parsing support should be largely complete. Next step
is to work on filling in sensible values (in IntegerLiteral/FloatLiteral).
llvm-svn: 39349
Submitted by:
Reviewed by:
First phase of parsing IntegerConstants. At the moment, all processing
is done in the Parser, not Sema. If necessary, this is easy to move.
Next steps:
- Convert well for strings to actual values (need to look @ APInt.h)
- Design the API between the Parser and Sema. Sema shouldn't have to be
concerned with any parsing issues...
llvm-svn: 39348
it to Sema/ASTStreamer (separating the lifetime of ASTContext from
the lifetime of Sema). One day it might be useful to consider creating
a context object implicitly if one isn't provided (using default arguments in
Sema's constructor). At this point, adding this convenience isn't necessary.
llvm-svn: 39346
the prerocessor will be available via ASTContext).
- Removed the public data member "PP" in ASTContext.
- Changed ASTContext's contructor to take TargetInfo/IdentifierTable explicitly.
- Implicitly create an ASTContext in Sema's constructor. This simplifies
the clients job (and makes ASTContext more private).
--As a side effect, added a "PrintStats" hook to Sema.
To support this level of encapsulation, ASTContext is always dynamically
allocated (by Sema). Previously, ASTContext was being allocated on the
stack. I don't believe this should be a performance issue (since ASTContext
is fairly course grain and tied to the lifetime of Sema currently).
llvm-svn: 39345
This set of changes includes:
- convert ExprCXX.h and add to the Xcode project file.
-- required adding CXXBoolLiteralExpr to StmtNodes.def.
-- required adding visitor support (decl/defn).
- make the class codes in StmtNodes.def explicit (to enable range checking).
-- this required changing all clients of the STMT macro.
- declare the instance data const.
llvm-svn: 39344
- adding enum constants & instance data to Stmt.
- adding classof() functions to all Stmt's.
- modifying contructors to pass the appropriate enum as an arg.
Also tightened up a couple "void *" declarations/casts for arrays.
llvm-svn: 39343
clang still compiled/linked/ran properly...simply a confusing name regression.
From now on I'll make sure I run "cvs diff" before committing any changes!
llvm-svn: 39342
of source code. For example:
$ clang INPUTS/carbon_h.c -arch i386 -arch ppc
prints:
...
/usr/lib/gcc/i686-apple-darwin8/4.0.1/include/mmintrin.h:51:3: note: use of a target-specific builtin function, source is not 'portable'
__builtin_ia32_emms ();
^
because carbon.h pulls in xmmintrin.h, and __builtin_ia32_emms isn't a builtin on ppc.
Though clang now supports target-specific builtins, the full table isn't implemented yet.
llvm-svn: 39328
whose decl objects are lazily created the first time they are referenced.
Builtin functions are described by the clang/AST/Builtins.def file, which
makes it easy to add new ones.
This is missing two important pieces:
1. Support for the rest of the gcc builtins.
2. Support for target-specific builtins (e.g. __builtin_ia32_emms).
Just adding this builtins reduces the number of implicit function definitions
by 6, reducing the # diagnostics from 550 to 544 when parsing carbon.h.
I need to add all the i386-specific ones to eliminate several hundred more.
ugh.
llvm-svn: 39327
SmallPtrSet data structure. This datastructure handles the 'nonsmall' case
quite gracefully, with an efficient exponentially probed hashtable. This is
important for handling global scope, which gets many thousands of decls (e.g.
every function and enum value). Of course the typical inner scopes are still
as efficient as ever.
On my mac pro, this speeds up parsing carbon.h from 0.59s to 0.168s (3.5x),
and there is still low hanging fruit :).
For reference, GCC on the same system takes 0.33s for -fsyntax-only.
llvm-svn: 39317
"obviously braindead" linear searches. reduces the number of slow
type lookups from 10K to 883 on carbon.h, speeding up parsing from 3.5 to
1.26s.
llvm-svn: 39312
a foldingset instead. This reduces the number of slow type lookups from
32K to 10K, which speeds up parsing of carbon.h from 11s to 3.5s.
llvm-svn: 39311
ASTContext::getTagDeclType by not having to do a linear search. With this,
parse time for carbon.h drops from 21.8s to 16.0s and # slow lookups drop from
83K to 63K.
llvm-svn: 39306
*** AST Context Stats:
30594 types total.
19 builtin types
3929 pointer types
308 array types
18883 function types with proto
8 function types with no proto
2988 typename (typedef) types
4459 tagged types
1476 struct types
80 union types
0 class types
2903 enum types
83298 slow type lookups
Next up, making type canonicalization not trivially silly.
llvm-svn: 39305
struct q { int a, a; };
with:
t.c:3:19: error: duplicate member 'a'
struct q { int a, a; };
^
t.c:3:16: error: previous definition is here
struct q { int a, a; };
^
llvm-svn: 39303
This emits these diagnostics:
t.c:4:14: error: redefinition of 'a'
enum foo22 { a, b };
^
t.c:3:5: error: previous definition is here
int a;
^
t.c:8:17: error: redefinition of enumerator 'b'
enum foo23 { c, b };
^
t.c:4:17: error: previous definition is here
enum foo22 { a, b };
^
4 diagnostics generated.
for:
int a;
enum foo22 { a, b };
enum foo23 { c, b };
llvm-svn: 39302
t.c:10:15: warning: 'bonk' may not be nested in a struct due to flexible array member
struct bink bonk;
^
t.c:13:14: error: 'struct bink' may not be used as an array element due to flexible array member
struct bink A[123];
^
for:
struct bink {
struct bink *a;
int X[]; // ok.
};
struct foo {
int A;
struct bink bonk;
};
struct bink A[123];
llvm-svn: 39296
struct bork {
int X[];
};
struct bink {
struct bink a;
int X[]; // ok.
};
to:
t.c:3:7: error: flexible array 'X' not allowed in otherwise empty struct
int X[];
^
t.c:7:15: error: field 'a' has incomplete type
struct bink a;
^
llvm-svn: 39295
like:
struct S { struct S {} X; };
with:
t.c:2:19: error: nested redefinition of 'struct'
struct S { struct S {} X; };
^
t.c:2:1: error: previous definition is here
struct S { struct S {} X; };
^
llvm-svn: 39292
struct blah * P;
union blah *P2;
we now emit:
t.c:2:1: error: redefinition of 'blah' with tag that does not match previous use
union blah *P2;
^
t.c:1:8: error: previous use is here
struct blah * P;
^
llvm-svn: 39275
- Add support for getting the namespace of a decl.
- Move type down the inheritance hierarchy, it doesn't make sense for a struct
to have a type, and typedef's use types in different ways than variables do.
- Add initial TagDecl/RecordDecl classes.
llvm-svn: 39267
Detect and emit errors when names are redefined in the same scope, e.g.
test/Parser/argument_redef.c, which now emits:
argument_redef.c:4:22: error: redefinition of 'A'
int foo(int A) { int A; }
^
argument_redef.c:4:13: error: previous definition is here
int foo(int A) { int A; }
^
llvm-svn: 39257
X = sizeof(int (void a));
X = sizeof(int (int, void));
X = sizeof(int (void, ...));
We now emit:
t.c:6:24: error: void argument may not have a name
X = sizeof(int (void a));
^
t.c:7:24: error: 'void' must be the first and only parameter if specified
X = sizeof(int (int, void));
^
t.c:8:19: error: 'void' must be the first and only parameter if specified
X = sizeof(int (void, ...));
^
And we pretty print this correctly (even though void isn't stored in the
arg list of the function type):
X = sizeof(int (void));
However, this approach will have to change to handle typedefs of void.
llvm-svn: 39235
declarator interface handles all alloc/dealloc issues related to the argument
list. Before the client had to alloc and Declarator did the dealloc.
llvm-svn: 39234
This lets us pretty print stuff like this:
void foo() {
int X;
X = sizeof(void (*(*)())());
X = sizeof(int(*)(int, float, ...));
X = sizeof(void (*(int arga, void (*argb)(double Y)))(void* Z));
as:
X = sizeof(void (*(*)())())
X = sizeof(int (*)(int, float, ...))
X = sizeof(void (*(int, void (*)(double)))(void *))
Ah the wonders of 'modern' C syntax!
llvm-svn: 39232
the info. Also, call Actions.ParseParamDeclaratorType instead of
Actions.ParseDeclarator for parameter type lists: we don't want declaration
objects created when parsing a function declarator, we just want type info.
llvm-svn: 39230
parameters: build an array of ParamInfo structures and pass it to the
declarator for safe keeping (it owns the list).
Next step: actually populate the arg array with useful stuff.
llvm-svn: 39229
things like:
t.c:4:10: error: invalid storage class specifier in function declarator
int foo2(auto int Aaslfkasdflkj, register B);
^
instead of:
t.c:4:19: error: invalid storage class specifier in function declarator
int foo2(auto int Aaslfkasdflkj, register B);
^
llvm-svn: 39224
of having a loose collection of function pointers. This also allows clients to
maintain state, and reduces the size of the Preprocessor.h interface.
llvm-svn: 39203
typedef int X;
int A() {
return X;
}
int B() {
return Y;
}
as:
/Users/sabre/test.c:5:10: error: unexpected type name 'X': expected expression
return X;
^
/Users/sabre/test.c:9:10: error: use of undeclared 'Y' value
return Y;
^
llvm-svn: 39192
thing properly. This allows us to print types like:
int (*A)[restrict static 4][6];
properly, in addition to representing them properly. :)
llvm-svn: 39178
fundamentally requires having an AST around, so move all sema to the
AST library. This is the first step, later steps will be needed to
clean up libast.
llvm-svn: 39150
case, speeding up parsing of this contrived example:
#define A {{}}
#define B A A A A A A A A A A
#define C B B B B B B B B B B
#define D C C C C C C C C C C
#define E D D D D D D D D D D
#define F E E E E E E E E E E
#define G F F F F F F F F F F
#define H G G G G G G G G G G
void foo() {
H
}
from 7.478s to 4.321s. GCC requires 8.2s.
llvm-svn: 39138
switch statements. Make break/continue check that they are inside of an
appropriate control-flow construct. This implements Parser/bad-control.c.
llvm-svn: 39136
filenames (and also '#pragma GCC dependency' of course). Now, assuming
no cleaning is needed, we can go all the way from lexing the filename to
doing filename lookups with no mallocs. This speeds up user PP time from
0.077 to 0.075s for Cocoa.h (2.6%).
llvm-svn: 39092
class instead. SmallString allows to code to avoid hitting malloc in
the normal case (or will, when some other stuff is converted over).
llvm-svn: 39084
declarations through the asm streamer. For a testcase like:
int G;
int H, I, *J;
int func() {}
'clang -parse-print-ast' prints:
Read top-level decl: G
Read top-level decl: H
Read top-level decl: I
Read top-level decl: J
Read top-level decl: func
llvm-svn: 38992
1. Allow argument list of a call to be empty.
2. Handle expressions in a statement context that start with an identifier.
These are nasty to handle due to the 'label:' production which requires
having two tokens of look-ahead, or by having tricky code (which we
choose to do).
llvm-svn: 38887
argument tokens into instead of a real vector. This avoids some malloc
traffic in common cases. In an "abusive macro expansion" testcase, this
reduced -Eonly time by 25%.
llvm-svn: 38757
is an intra-lexer property, not a inter-lexer property, so it makes sense
for it to be define here. It also makes no sense for macros, and allows us
to define it more carefully in the header.
While I'm at it, improve comments and structuring in Lexer.h
llvm-svn: 38701
that are lexed are made up of tokens, so the calls are just ugly and redundant.
Hook up the MIOpt for the #if case. PPCExpressions doesn't currently implement
the hook though, so we still don't handle #if !defined(X) with the MIOpt.
llvm-svn: 38649
came from a macro expansion, this allows us to keep track of both where the
character data came from and where the logical position of the token is (at
the expansion site). This implements Preprocessor/indent_macro.c, and
reduces the number of cpp iostream -E diffs vs GCC from 2589 to 2297.
llvm-svn: 38557
Now, instead of keeping a pointer to the start of the token in memory, we keep the
start of the token as a SourceLocation node. This means that each LexerToken knows
the full include stack it was created with, and means that the LexerToken isn't
reliant on a "CurLexer" member to be around (lexer tokens would previously go out of
scope when their lexers were deallocated).
This simplifies several things, and forces good cleanup elsewhere. Now the
Preprocessor is the one that knows how to dump tokens/macros and is the one that
knows how to get the spelling of a token (it has all the context).
llvm-svn: 38551
Steve Naroff