[xml] Probably should not look for/use shl_load if dlopen is found (2.6.17)

[Rick Jones <rick jones2 hp com>]

On some HP-UX systems, both dlopen and shl_load are present.  The 2.6.17 
configure script seems to look for both of them and then xmlmodule.c does not 
prefer one over the other.

When both shl_load and dlopen are present, we get duplicate definitions of the 
routines and that makes the compiler rather unhappy :)

I'm not sure which would be better - only have configure look for shl_load if it 
does not find dlopen, or modify the #ifdefs in xmlmodule.c such that the 
shl_load stuff is nested in an #else from the HAVE_DLOPEN.  This patch does the 
latter - as much because I didn't have autoconf installed on the HP-UX system as 
anything else :)

# diff -c  xmlmodule.c.orig xmlmodule.c
*** xmlmodule.c.orig    Wed Mar  9 11:47:20 2005
--- xmlmodule.c Wed Mar  9 11:47:23 2005
***************
*** 237,243 ****
      return 0;
  }

! #endif /* HAVE_DLOPEN */

  #ifdef HAVE_SHLLOAD             /* HAVE_SHLLOAD */
  #ifdef HAVE_DL_H
--- 237,243 ----
      return 0;
  }

! #else

  #ifdef HAVE_SHLLOAD             /* HAVE_SHLLOAD */
  #ifdef HAVE_DL_H
***************
*** 285,290 ****
--- 285,291 ----
  }

  #endif /* HAVE_SHLLOAD */
+ #endif /* HAVE_DLOPEN */

  #ifdef _WIN32


that takes-care of the compilation errors on 11.23 IPF, and the code still 
compiles happily on 11.11 PA-RISC.  I do not as yet have access to an 11.23 
PA_RISC system, but I suspect it would still be happy there.  This is all a 
default 32-bit compilation.

rick jones