Re: RFC: COMDAT group names become anonymouse local symbols

[Dave Korn <dave dot korn dot cygwin at gmail dot com>]

On 24/10/2010 05:34, Mark Mitchell wrote:
> On 10/23/2010 8:34 PM, Hans-Peter Nilsson wrote:
> 
>>> Does that mean adding a new entry to struct format_ops?
>> That's it.  I didn't say, because I'm not familiar too with this
>> code. :)  I only discovered it and fixed a few bugs eons ago.
> 
> I'm traveling tomorrow, so might not get a chance to finish this up.  (I
> need to test a few different configurations to convince myself I've
> filled in the various blanks correctly.)  Here's a preliminary patch
> which seems to pass the LD testsuite for cris-elf.  Does it work for you?

  Hi, I haven't been following this thread so far, but this hunk:

> --- 1941,1948 ----
>     0,	/* sec_sym_ok_for_reloc */
>     coff_pop_insert,
>     0,	/* ecoff_set_ext */
>     coff_obj_read_begin_hook,
>     coff_obj_symbol_new_hook,
> !   coff_obj_symbol_clone_hook,
> !   coff_obj_adjust_symtab
>   };

... breaks i686-pc-cygwin, and presumably all other COFF targets:

> /gnu/binutils/src/gas/config/obj-coff.c:1947: error: 'coff_obj_adjust_symtab'
> undeclared here (not in a function)
> cc1: warnings being treated as errors
> /gnu/binutils/src/gas/config/obj-coff.c:1948: error: missing initializer
> /gnu/binutils/src/gas/config/obj-coff.c:1948: error: (near initialization for 
> 'coff_format_ops.adjust_symtab')
> make[4]: *** [obj-coff.o] Error 1

  I tried changing the struct entry to 'coff_adjust_symtab' and the build
completes, but so would putting a null there, like in the ecoff version...
which is right?

    cheers,
      DaveK