This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Bad overload resolution


Using the attached file, gcc gives me the following (incorrect) output:
[bcurrie@nzlc01 ~]
$ gcc -O2 -S oload-bug.cc -v
Reading specs from
/home/users/bcurrie/lib/gcc-lib/hppa1.1-stratus-sysv4/egcs-2.92.00/specs
gcc version egcs-2.92.00 19980715 (gcc2 ss-980609 experimental)
 /home/users/bcurrie/lib/gcc-lib/hppa1.1-stratus-sysv4/egcs-2.92.00/cpp
-lang-c++ -v -undef -D__GNUC__=2 -D__GNUG__=2 -D__cplusplus
-D__GNUC_MINOR__=92 -Dhppa -Dunix -D_FTX -D__hppa__ -D__unix__ -D_FTX
-D__hppa -D__unix -Asystem(unix) -Acpu(hppa) -Amachine(hppa)
-D__EXCEPTIONS -D__OPTIMIZE__ -D_PA_RISC1_1 oload-bug.cc
/var/tmp/cc5MPBB8.ii
GNU CPP version egcs-2.92.00 19980715 (gcc2 ss-980609 experimental)
(hppa)
#include "..." search starts here:
#include <...> search starts here:
 /home/users/bcurrie/include/g++
 /usr/local/include
 /home/users/bcurrie/hppa1.1-stratus-sysv4/include

/home/users/bcurrie/lib/gcc-lib/hppa1.1-stratus-sysv4/egcs-2.92.00/include
 /usr/include
End of search list.

/home/users/bcurrie/lib/gcc-lib/hppa1.1-stratus-sysv4/egcs-2.92.00/cc1plus
/var/tmp/cc5MPBB8.ii -quiet -dumpbase oload-bug.cc -O2 -version -o
oload-bug.s
GNU C++ version egcs-2.92.00 19980715 (gcc2 ss-980609 experimental)
(hppa1.1-stratus-sysv4) compiled by GNU C version 2.7.2.1.
oload-bug.cc: In method `foobar::foobar()':
oload-bug.cc:21: warning: assuming & on `foobar::callback()'
oload-bug.cc:21: warning: converting from `void (foobar::*)()' to `void
(*)(void *)'

If I reverse the two callback definitions, everything is fine and the
correct address is passed to `func'.

In the file where I first encountered this problem, I actually got an
error, but that was probably because `func' was overloaded as well, but
reversing the declarations of the callback function fixed it.  I also
checked the 19980730 version and had the same problem.

I take it that overloaded function resoltion for taking the address of a
function is not meant to be dependent on the order of definitions.  This
has always worked for me before (<=2.7.2.1, first time I tried building
this particular program with egcs).

Bill
-- 
Leave others their otherness
void func(void (*)(void *), void *);

class foobar {
    static void callback(void *);
    void callback();
  public:
    foobar();
};

void foobar::callback()
{
}

void foobar::callback(void *obj)
{
    ((foobar*)obj)->callback();
}

foobar::foobar()
{
    func(callback,this);
}

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]