[GCC-XML]typedef and array in gccxml

Brad King brad.king at kitware.com
Wed Oct 30 17:41:48 EST 2002


> typedef unsigned char logical;
> void check_datatypes(logical alogic){};
> <Argument name="aLogical" type="_6" />
> <FundamentalType name="unsigned char" id="_6" />
> not exactly what I expected since I loss synonym information in my
> function.

This is an unfortunate consequence of the way GCC stores the argument
information internally.  I have found no way to recover the original
typedef in all cases.  There is a "DECL_ARGUMENT_AS_WRITTEN" macro in
GCC's source that is supposed to extract the original type of the argument
from a PARM_DECL (a function argument).  However, this method does not
seem to work well all the time.  I'm guessing that I've mis-read the
documentation, but I don't know if there is anything I can do about it.

> /for  double  origin[3]/
>  gccxml falsly interprets as
> /<Argument name="origin" type="_129" />
> <PointerType id="_129" type="_121" />
> <FundamentalType name="double" id="_121" /> /
> no array information here.

Again, this is the result of GCC's internal storage of the types.
However, in this case it is correct.  The C++ standard specifies in 13.1/3
that in a function argument, the outer-most array bounds are ignored and
treated instead as a pointer type.  You will find that array information
is correctly reported for the inner-bounds of a multi-dimensional array

The following are equivalent.  Each specifies that the argument is a
pointer to an array of 3 integers.  Whether that array of 3 integers is
the first of several in an array of arrays is up to the function body.

void f(int[3][3]);
void f(int [][3]);
void f(int(*)[3]);

You can also try running GCC-XML on this code to see that array types are
correctly reported:

typedef int A[3];


More information about the gccxml mailing list