Re: A COFF patch

Re: A COFF patch

[Nick Clifton]

HI Ian,

: I checked in the appropriate patch.  Nick omitted updating
: coff-m88k.c.

Oops!  Thanks for catching that.

Cheers
	Nick

Re: A COFF patch

[Ian Lance Taylor]

   Date: Mon, 19 Jul 1999 11:44:40 -0700 (PDT)
   From: hjl@varesearch.com (H.J. Lu)

   Just configure binutils with "--enable-targets=all" on Linux/x86
   and watch for "make check". Many tests failed on me.

I checked in the appropriate patch.  Nick omitted updating
coff-m88k.c.

Ian

Re: A COFF patch

[H.J. Lu]

> 
>    Date: Mon, 19 Jul 1999 10:29:37 -0700 (PDT)
>    From: hjl@varesearch.com (H.J. Lu)
> 
>    I don't know what you did to COFF. It seems that your change breaks
>    "configure --enable-targets=all". This patch seems to fix it. It may
>    not cover everything nor the best fix. But it is good enough for me
>    to pass "make check" when I use "configure --enable-targets=all" on
>    Linux/x86.
> 
>    Mon Jul 19 10:24:55 1998  H.J. Lu  (hjl@gnu.org)
> 
> 	   * coffgen.c (coff_object_p): Check if we are fed with a COFF
> 	   file.
> 
> This patch is definitely wrong.  Nick, please don't check it in.
> 
> It is simply wrong for coff_object_p to be called with abfd->xvec set
> to any bfd_target structure which has a NULL backend_data field.  We
> need to understand why that is happening.  Nick added a field to
> bfd_target; perhaps the problem is simply that he somehow forgot to
> update one of the many structure initializations.
> 

Just configure binutils with "--enable-targets=all" on Linux/x86
and watch for "make check". Many tests failed on me.


-- 
H.J. Lu (hjl@gnu.org)

Re: A COFF patch

[Ian Lance Taylor]

   Date: Mon, 19 Jul 1999 10:29:37 -0700 (PDT)
   From: hjl@varesearch.com (H.J. Lu)

   I don't know what you did to COFF. It seems that your change breaks
   "configure --enable-targets=all". This patch seems to fix it. It may
   not cover everything nor the best fix. But it is good enough for me
   to pass "make check" when I use "configure --enable-targets=all" on
   Linux/x86.

   Mon Jul 19 10:24:55 1998  H.J. Lu  (hjl@gnu.org)

	   * coffgen.c (coff_object_p): Check if we are fed with a COFF
	   file.

This patch is definitely wrong.  Nick, please don't check it in.

It is simply wrong for coff_object_p to be called with abfd->xvec set
to any bfd_target structure which has a NULL backend_data field.  We
need to understand why that is happening.  Nick added a field to
bfd_target; perhaps the problem is simply that he somehow forgot to
update one of the many structure initializations.

Ian

A COFF patch

[H.J. Lu]

Hi, Nick,

I don't know what you did to COFF. It seems that your change breaks
"configure --enable-targets=all". This patch seems to fix it. It may
not cover everything nor the best fix. But it is good enough for me
to pass "make check" when I use "configure --enable-targets=all" on
Linux/x86.

Thanks.

-- 
H.J. Lu (hjl@gnu.org)
---
Mon Jul 19 10:24:55 1998  H.J. Lu  (hjl@gnu.org)

	* coffgen.c (coff_object_p): Check if we are fed with a COFF
	file.

Index: coffgen.c
===================================================================
RCS file: /work/cvs/gnu/binutils/bfd/coffgen.c,v
retrieving revision 1.1.1.2
diff -u -p -r1.1.1.2 coffgen.c
--- coffgen.c	1999/07/12 17:13:25	1.1.1.2
+++ coffgen.c	1999/07/19 17:21:51
@@ -254,6 +254,11 @@ coff_object_p (abfd)
   struct internal_filehdr internal_f;
   struct internal_aouthdr internal_a;
 
+  /* We may be called as bfd_check_format. We have to be prepared to
+     deal with everything. */
+  if (!coff_backend_info(abfd))
+    return 0;
+
   /* figure out how much to read */
   filhsz = bfd_coff_filhsz (abfd);
   aoutsz = bfd_coff_aoutsz (abfd);