sysdeps/powerpc/libgcc-compat.c problem (glibc-2-2-branch)

[Peter Bergner]

Ulrich,

We have the following code snipits on the CVS mainline:

include/libc-symbols.h:
    /* Handling on non-exported internal names.  We have to do this only
       for shared code.  */
    #ifdef SHARED
    # define INTUSE(name) name##_internal
    [snip]
    #else
    # define INTUSE(name) name
    [snip]
    #endif


sysdeps/powerpc/libgcc-compat.c:
    extern int64_t __ashldi3 (int64_t, int32_t);
    int64_t INTUSE (__ashldi3) (int64_t u, int32_t b)
    {
      return __ashldi3 (u, b);
    }
    symbol_version (INTUSE (__ashldi3), __ashldi3, GLIBC_2.0);


On the glibc-2-2-branch, you removed the use of the INTUSE macro
from sysdeps/powerpc/libgcc-compat.c which gives us:

    extern int64_t __ashldi3 (int64_t, int32_t);
    int64_t __ashldi3 (int64_t u, int32_t b)
    {
      return __ashldi3 (u, b);
    }
    symbol_version (__ashldi3, __ashldi3, GLIBC_2.0);


I'm assuming you didn't mean to create infinite recursion and want something
closer to what CVS mainline gets (after macro expansion):

    extern int64_t __ashldi3 (int64_t, int32_t);
    int64_t __ashldi3_internal (int64_t u, int32_t b)
    {
      return __ashldi3 (u, b);
    }
    symbol_version (__ashldi3_internal, __ashldi3, GLIBC_2.0);

Ditto for the other routines in that file.


Peter

--
Peter Bergner
Linux PPC64 & GLIBC Kernel Development
IBM Rochester, MN
bergner@vnet.ibm.com