Re: binutils-2.9 objcopy fails (fwd)

Re: binutils-2.9 objcopy fails (fwd)

[Joel Sherrill]

I went ahead and asked the obvious questions on the m68k-rtems objcopy
failure.  I hope it helps.


---------- Forwarded message ----------
Date: Tue, 14 Apr 98 14:09:46 -0600
From: Eric Norum <eric@skatter.usask.ca>
To: Joel Sherrill <joel@oarcorp.com>
Subject: Re: binutils-2.9 objcopy fails

You wrote:
> I have forwarded this to the binutils list. What version of
> binutils did you upgrade from?

Thanks.  Perhaps you could forward this, too?
It's been so long since I had to make a 32-bit bootstrap prom that I  
don't remember which version I used to use. :-(

>
> I have breakages in powerpc-rtems and mips64orion-rtems. There is
> some known shared library problem. Are you building with
> --enabled-shared?

I just built with whatever `bit' does.

I'm pretty sure that I'm not trying to build anything with shared  
libraries.

Just for the heck of it, I tried running this under gdb.  I found  
that the text segment of prom.exe seems to be copied with no problem,  
but the data segment fails:
Breakpoint 3, copy_section (ibfd=0xe61a0, isection=0xe6994,  
obfdarg=0xe9220) at  
../../src/binutils/../../src/binutils/objcopy.c:1287
1287              if (!bfd_set_section_size (obfd, osection, size))
(gdb) step
1288                nonfatal (bfd_get_filename (obfd));
(gdb) print size
$5 = 1900


The fact that this fails the second time through is not too  
surprising, given the code at the beginning of  
bfd_set_section_size........

boolean
bfd_set_section_size (abfd, ptr, val)
     bfd *abfd;
     sec_ptr ptr;
     bfd_size_type val;
{
  /* Once you've started writing to any section you cannot create or  
change
     the size of any others. */

  if (abfd->output_has_begun)
    {
      bfd_set_error (bfd_error_invalid_operation);
      return false;
    }

---
Eric Norum                                 eric@skatter.usask.ca
Saskatchewan Accelerator Laboratory        Phone: (306) 966-6308
University of Saskatchewan                 FAX:   (306) 966-6058
Saskatoon, Canada.

Re: binutils-2.9 objcopy fails (fwd)

[Ian Lance Taylor]

   Date: Tue, 14 Apr 1998 14:23:14 -0500 (CDT)
   From: Joel Sherrill <joel@OARcorp.com>

   ---------- Forwarded message ----------
   Date: Tue, 14 Apr 98 13:20:47 -0600
   From: Eric Norum <eric@skatter.USask.Ca>

   I downloaded and installed the latest binutils.  Now I can't  
   generate prom images any more.

   Background:
   I've got two types of 68360 systems here.  One uses a single 8-bit  
   bootstrap prom and the other uses 4 8-bit bootstrap proms as a 32-bit  
   prom.  I use a script to generate the S-record images for my prom  
   burner.

   Here are the salient lines from the script:

   m68k-rtems-objcopy --output-target=srec prom.exe prom.hex

   for byte in 0 1 2 3
   do
	   m68k-rtems-objcopy --interleave=4 --byte=$byte  
   --output-target=srec prom.exe prom$byte.hex
   done


   Problem:
   The first objcopy works fine to build the 8-bit bootstrap prom  
   image, but when I try to split the executable into the 4 images for  
   the 32-bit prom, I get:

   + m68k-rtems-objcopy --output-target=srec prom.exe prom.hex
   + m68k-rtems-objcopy --interleave=4 --byte=0 --output-target=srec  
   prom.exe prom0.hex
   m68k-rtems-objcopy: prom0.hex: Invalid operation


I can recreate this problem, but what's odd is that all of the code is
old.  What version of the binutils were you using before?  I can
recreate the problem using an binutils 2.6 I have lying around.

The problem I'm seeing is this call in copy_section in
binutils/objcopy.c:

              /* The section has gotten smaller. */
          if (!bfd_set_section_size (obfd, osection, size))
            nonfatal (bfd_get_filename (obfd));

This is trying to set the section size after objcopy has already
started to write to the sections.  This needs to be handled
differently.

Ian

binutils-2.9 objcopy fails (fwd)

[Joel Sherrill]

A failure in m68k-rtems objcopy.

--joel


---------- Forwarded message ----------
Date: Tue, 14 Apr 98 13:20:47 -0600
From: Eric Norum <eric@skatter.USask.Ca>
To: rtems-snapshots@oarcorp.com
Subject: binutils-2.9 objcopy fails

I downloaded and installed the latest binutils.  Now I can't  
generate prom images any more.

Background:
I've got two types of 68360 systems here.  One uses a single 8-bit  
bootstrap prom and the other uses 4 8-bit bootstrap proms as a 32-bit  
prom.  I use a script to generate the S-record images for my prom  
burner.

Here are the salient lines from the script:

m68k-rtems-objcopy --output-target=srec prom.exe prom.hex

for byte in 0 1 2 3
do
        m68k-rtems-objcopy --interleave=4 --byte=$byte  
--output-target=srec prom.exe prom$byte.hex
done


Problem:
The first objcopy works fine to build the 8-bit bootstrap prom  
image, but when I try to split the executable into the 4 images for  
the 32-bit prom, I get:

+ m68k-rtems-objcopy --output-target=srec prom.exe prom.hex
+ m68k-rtems-objcopy --interleave=4 --byte=0 --output-target=srec  
prom.exe prom0.hex
m68k-rtems-objcopy: prom0.hex: Invalid operation


Any ideas?
---
Eric Norum                                 eric@skatter.usask.ca
Saskatchewan Accelerator Laboratory        Phone: (306) 966-6308
University of Saskatchewan                 FAX:   (306) 966-6058
Saskatoon, Canada.