Re: elf32.em

[Jeffrey A Law <law@cygnus.com>]

  In message < 19991027234806.10154.qmail@daffy.airs.com >you write:
  >    Date: Wed, 27 Oct 1999 18:48:30 +1000
  >    From: Geoff Keating <geoffk@ozemail.com.au>
  > 
  >    In elf32.em, there's this code:
  > 
  > 	FIXME: This approach--using bfd_elf_set_dt_needed_name--is not
  > 	very pretty.  I haven't been able to think of anything that is
  > 	pretty, though.  */
  >      if (bfd_check_format (entry->the_bfd, bfd_object)
  > 	 && (entry->the_bfd->flags & DYNAMIC) != 0)
  >        {
  > 	 char *filname, *needed_name;
  > 
  > 	 ASSERT (entry->is_archive && entry->search_dirs_flag);
  > 
  > 	 /* Rather than duplicating the logic above.  Just use the
  > 	    filename we recorded earlier.o
  > 
  > 	    First strip off everything before the last '/'.  */
  > 	 filename = strrchr (entry->filename, '/');
  > 
  >    Shouldn't the variable be called 'filename'?
  > 
  > Actually, I think the uses of it should be filname, to avoid
  > shadowing.  Or we should just remove the declaration of filname, and
  > use the existing filename local.
  > 
  > Jeff, I'm not sure it's a good idea to try for the .sl extension on
  > every system type, given that it is only meaningful on HP/UX.  If
  > there is a long search path, it is a waste of time.  It is also
  > incorrect in principle.  You can test for a particular emulation using
  > #ifdef TARGET_IS_xxx.
I found out what happened with this.

Uli never checked in the version that conditionalized the hpux check to
the binutils repository.  Ugh.

Instead of using a TARGET_IS_xxx, he has a macro which specifies an alternate
shared library extension to check.  That seems to make more sense to me as
other targets could potentially use it if necessary.

jeff