🚀 go-pugleaf

The following program

-------cut here---------
#include <new.h>

class foo
{
public:
  int baz;
  foo(int b): baz(b) {}
};

main()
{
  foo* foop = (foo*) malloc(sizeof(foo));
  foop = new {foop} foo(4);
}
-------cut here---------

where new.h comes from libg++-2.0,
does not link successfully.

-------cut here---------
cd /marsh/comp212/.jnkk/
c++ -g -Wall -v bug.cc
Reading specs from /marsh/comp212/lib/gcc-lib/sun4/2.1/specs
gcc version 2.1
 /marsh/comp212/lib/gcc-lib/sun4/2.1/cpp -lang-c++ -v -undef -D__GNUC__=2 -D__GNUG__=2 -D__cplusplus -Dsparc -Dsun -Dunix -D__sparc__ -D__sun__ -D__unix__ -D__sparc -D__sun -D__unix -g -Wall bug.cc /usr/tmp/cca14672.i
GNU CPP version 2.1 (sparc)
 /marsh/comp212/lib/gcc-lib/sun4/2.1/cc1plus /usr/tmp/cca14672.i -quiet -dumpbase bug.cc -g -Wall -version -o /usr/tmp/cca14672.s
GNU C++ version 2.1 (sparc) compiled by GNU C version 2.1.
 as -o bug.o /usr/tmp/cca14672.s
 /marsh/comp212/lib/gcc-lib/sun4/2.1/ld -e start -dc -dp /lib/crt0.o -L/marsh/comp212/lib/gcc-lib/sun4/2.1/ -L/marsh/comp212/lib/ bug.o -lg++ -lm -lgcc -lc -lg -lgcc
ld: Undefined symbol 
   ___nw__FiPv 
collect: /usr/bin/ld returned 2 exit status

Compilation exited abnormally with code 1 at Sun Apr 12 01:05:36
-------cut here---------

However, when I comment out the #pragma interface line of new.h, it
does link successfully.  The file new.h includes an inline static
definition of the function that is claimed to be undefined.

I am not making the same complaint as the gentleman who complained
that nm showed a particular symbol to be undefined.  I am complaining
that inlining does not happen in the presence of #pragma interface, at
least for operator new with placement syntax.

That is why I'm sending this to g++.bug, not to libg++.bug.

Sparc 2.  Sunos 4.1.  Gcc 2.1, with Tiemann and Stump patches.

Doug Moore
(dougm@cs.rice.edu)