We need to convert C++ integer types to C# types using the TargetInfo, as the
type bitwidths depend on the target architecture.
This fixes the main 64bit issue, which is that on 64bit Linux the current code
converts C++'s long to System.Int32, instead of System.Int64.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
We need Driver.TargeInfo in CSharpTypePrinter so that we can find out the
widths of the integer types.
This patch changes CSharpTypePrinter to get the Driver reference in its
constructor. As all the other constructor parameters can be found from the
Driver reference, we can remove all the other parameters.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
As enum PrimitiveType is supposed to represent C++ types, having Int32
and UInt32 there is not correct.
This patch renames those values to Int and UInt.
This doesn't change the behavior, but makes it clearer that the types
for 'int' and 'unsigned int' are not necessarily 32 bits.
As enum PrimitiveType is supposed to represent C++ types, having Int64
and UInt64 there is not correct.
This patch renames those values to LongLong and ULongLong.
This doesn't change the behavior, but makes it clearer that the types
for 'long long' and 'unsigned long long' are not necessarily 64 bits.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
As enum PrimitiveType is supposed to represent C++ types, having Int16
and UInt16 there is not correct.
This patch renames those values to Short and UShort.
This doesn't change the behavior, but makes it clearer that the types
for 'short' and 'unsigned short' are not necessarily 16 bits.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
As enum PrimitiveType is supposed to represent C++ types, having Int8
and UInt8 there is not correct.
This patch removes those values. As Int8 and UInt8 were set to be
equivalent to Char and UChar, we can just change all references to Int8
and UInt8 to Char and UChar.
This doesn't change the behavior, but makes it clearer that the types
for 'char' and 'unsigned char' are not necessarily 8 bits.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
CppTypePrinter outputs C++ code, so the proper type to be written for
Long and ULong are "long" and "unsigned long", not "int" and "unsigned
int".
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
CLITypePrinter outputs C++ code, so the proper type to be written for
Long and ULong are "long" and "unsigned long", not "int" and "unsigned
int".
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
The enum PrimitiveType represents, according to the comment, "the C++
built-in types.". However, it does not contain values for C++'s "long"
or "unsigned long", and the parser converts long to int. This is not
correct, and is a problem with 64bit support.
This patch adds Long and ULong values to PrimitiveType, and adds those
to all the switch-cases all around the code. For now Long and ULong are
handled exactly like Int and UInt, so this patch does not change
behavior.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
C++ types cannot be converted to C# types without knowing details of the
target (e.g. what is the size of "long") and
PrimitiveTypeExtensions.ConvertToType does not handle that correctly.
However, ConvertToType is not used anywhere, so let's just remove it so
that no one will use the method.
We could also fix the method, but it's an extension method for
PrimitiveType, and I don't think such an extension method should be used
for the task, as you would also need to pass it information about the
target system.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Now that the warnings have been fixed, we no longer need -fpermissive to
compile CppParser with gcc.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Fix the following warning by renaming ParserResult::Parser to
CodeParser.
In file included from ../../../src/CppParser/Parser.h:16:0,
from ../../../src/CppParser/Comments.cpp:8:
../../../src/CppParser/CppParser.h:86:13: warning: declaration of ‘CppSharp::CppParser::Parser* CppSharp::CppParser::ParserResult::Parser’ [-fpermissive]
Parser* Parser;
^
../../../src/CppParser/CppParser.h:73:8: warning: changes meaning of ‘Parser’ from ‘struct CppSharp::CppParser::Parser’ [-fpermissive]
struct Parser;
^
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Fix the following warning by renaming RawComment::FullComment to
FullCommentBlock.
In file included from ../../../src/CppParser/AST.cpp:8:0:
../../../src/CppParser/AST.h:783:18: warning: declaration of ‘CppSharp::CppParser::AST::FullComment* CppSharp::CppParser::AST::RawComment::FullComment’ [-fpermissive]
FullComment* FullComment;
^
../../../src/CppParser/AST.h:760:15: warning: changes meaning of ‘FullComment’ from ‘struct CppSharp::CppParser::AST::FullComment’ [-fpermissive]
struct CS_API FullComment : public Comment
^
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Fix the following warning by renaming RawComment::RawCommentKind to
Kind.
In file included from ../../../src/CppParser/AST.cpp:8:0:
../../../src/CppParser/AST.h:780:20: warning: declaration of ‘CppSharp::CppParser::AST::RawCommentKind CppSharp::CppParser::AST::RawComment::RawCommentKind’ [-fpermissive]
RawCommentKind RawCommentKind;
^
../../../src/CppParser/AST.h:765:13: warning: changes meaning of ‘RawCommentKind’ from ‘enum class CppSharp::CppParser::AST::RawCommentKind’ [-fpermissive]
enum struct RawCommentKind
^
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
There are multiple places in the parser which do something similar to
this:
Foo *ptr = 0;
if (...) {
Foo val = ...;
ptr = &val;
}
func(ptr);
In other words, the code takes a pointer to a local variable, and the
pointer is used even after the local variable's scope has ended.
This causes a crash on Linux. For some reason this works fine on Windows
+ Visual studio.
This patchs moves the variable ("val" in the above example) to outer
scopes, so that its life time is extended to cover the use.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Fix trivial switch/case compile warnings like the one below by adding default
case where suitable, or adding the missing cases when that makes more sense.
enumerator 'LastPreprocessingDirective' in switch of enum 'clang::PreprocessedEntity::EntityKind' is not handled
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Fix trivial switch/case compile warnings like the one below by adding default
case where suitable, or adding the missing cases when that makes more sense.
enumerator 'LastPreprocessingDirective' in switch of enum 'clang::PreprocessedEntity::EntityKind' is not handled
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Use forward declarations to avoid the need to include lots of clang headers.
This drops the compilation time for CppParser from 17.8 seconds to 14.1 seconds
for me.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
TestDestructors test cannot be run twice as a static variable holds data from
the first run.
Add an init function to set the static variable to a known state.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
DriverOptions.Is32Bit is used to decide whether to use 4 or 8 byte
pointer size in VTable calculations.
Instead of having a settable Is32Bit property, use
TargetInfo.PointerWidth for this. This allows us to remove the whole
Is32Bit property, and the IS_64_BIT define.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@iki.fi>
Tom Spilman
triton
Tomi Valkeinen
Elias Holzer