Re: Re C4x patch

[Ian Lance Taylor <ian at zembu dot com>]

   Date: Thu, 23 Dec 1999 12:33:42 -0800
   From: Nick Clifton <nickc@cygnus.com>

     * This patch (to bfd/coffcode.h) would not apply.  I could not
       locate this code anywhere, so I have left it out.

   *** 1068,1078 ****
   --- 1068,1080 ----
	  that is also how gdb locates the section.
	  We need to handle the .ctors and .dtors sections similarly, to
	  avoid introducing null words in the tables.  */
   + #ifndef C4XMAGIC
       if (COFF_DEFAULT_SECTION_ALIGNMENT_POWER > 2
	   && (strncmp (section->name, ".stab", 5) == 0
	     || strcmp (section->name, ".ctors") == 0
	     || strcmp (section->name, ".dtors") == 0))
	 section->alignment_power = 2;
   + #endif

This has been replaced by the coff_set_custom_section_alignment
support, in which the target file defines
COFF_SECTION_ALIGNMENT_ENTRIES if necessary.

The patch above is almost certainly incorrect.  It's hard for me to
imagine any target for which it would not be appropriate to set the
alignment of those sections.

Nick, as you know, please be extremely careful about any patches to
any files which are not specific to the C4x.

     * Why did you make the code to handle s->fix_line at line 745 of
       bfd/coffgen.c conditional only on RS6000COFF_C ?

Any such change would be incorrect, as coffgen.c is compiled for all
COFF targets.  It is incorrect to test a target specific macro like
RS6000COFF_C in that file.

It so happens that s->fix_line will only be true when working with an
RS/6000 COFF file.  However, it is incorrect to add a #ifdef test to
coffgen.c.  Not only is it the wrong approach, it won't even work
correctly--it will break RS/6000 COFF support.

     * This patch (to gas/config/obj-coff.c) would not apply. I could not
       locate the correct place to apply this patch, so I left it out.

This code is in c_symbol_merge.  I believe it is trying to fix a bug
which was fixed by this change:

Mon Jul 13 13:55:42 1998  Ian Lance Taylor  <ian@cygnus.com>

	* config/obj-coff.c (c_symbol_merge): Correct number of bytes when
	copying aux information.



     * Why does your patch remove a test on flag_m68k_mri at lines 295
       and 509 of gas/expr.c ?

I believe any such change would be incorrect.

Ian