From mboxrd@z Thu Jan  1 00:00:00 1970
From: "H . J . Lu" <hjl@lucon.org>
To: Nick Clifton <nickc@cambridge.redhat.com>
Cc: binutils@sources.redhat.com, Ian Lance Taylor <ian@zembu.com>
Subject: Re: elf_link_hash_entry vs generic_link_hash_entry
Date: Fri, 24 Aug 2001 10:02:00 -0000
Message-id: <20010824100230.A5440@lucon.org>
References: <si66bh2zdl.fsf@daffy.airs.com> <m3y9ocbkyw.fsf_-_@north-pole.nickc.cambridge.redhat.com> <20010823092140.A15101@lucon.org> <20010823113656.A17692@lucon.org> <20010823120958.A18304@lucon.org> <m3lmk9fmgc.fsf@north-pole.nickc.cambridge.redhat.com>
X-SW-Source: 2001-08/msg00567.html

On Fri, Aug 24, 2001 at 05:32:19PM +0100, Nick Clifton wrote:
> 
> I had to make one change though - I reverted the definition of
> elf_hash_table() back to its original version, since it cannot be
> allowed to return NULL.  (It is used in too many places as the first
> argument to elf_link_hash_lookup).  Instead I created a new macro
> called 'is_elf_hash_table' and I added code to check it in
> elf_link_add_object_symbols(), like this.
> 

I don't like it. I prefer elf_hash_table

1. Returns NULL on generic hash table

#define elf_hash_table(p)	\
  ((struct elf_link_hash_table *) \
     (((p)->hash->type == bfd_link_elf_hash_table) ? (p)->hash : NULL))

so that ld gets a core dump. Or

2. Aborts on generic hash table

#define elf_hash_table(p)	\
  ((struct elf_link_hash_table *) \
     (((p)->hash->type == bfd_link_elf_hash_table)
      ? (p)->hash : (abort (), NULL)))

so that we can catch all the bogus code.


H.J.