[Bug binutils/23947] New: objcopy refuses to copy !(SEC_LOAD|SEC_ALLOC) sections

[ppluzhnikov at google dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=23947

            Bug ID: 23947
           Summary: objcopy refuses to copy !(SEC_LOAD|SEC_ALLOC) sections
           Product: binutils
           Version: 2.32 (HEAD)
            Status: NEW
          Severity: normal
          Priority: P2
         Component: binutils
          Assignee: unassigned at sourceware dot org
          Reporter: ppluzhnikov at google dot com
  Target Milestone: ---

On Fedora, /usr/bin/ls has a mini-symbols .gnu_debugdata section:

readelf -WS /usr/bin/ls 
There are 30 section headers, starting at offset 0x26148:
...
  [27] .gnu_debuglink    PROGBITS        0000000000000000 025138 000020 00     
0   0  4

I wanted to examine the contents of this section, so I did the usual:

  objcopy -O binary -j.gnu_debuglink /usr/bin/ls /tmp/ls.mini.xz

but that produced an empty file. I had to resort to manually computing offset
and size, and doing dd to extract the contents.

AFAICT, this is happening because of the following check in bfd/binary.c:

284       /* We don't want to output anything for a section that is neither
285          loaded nor allocated.  The contents of such a section are not
286          meaningful in the binary format.  */
287       if ((sec->flags & (SEC_LOAD | SEC_ALLOC)) == 0)
288         return TRUE;
289       if ((sec->flags & SEC_NEVER_LOAD) != 0)
290         return TRUE;
291     
292       return _bfd_generic_set_section_contents (abfd, sec, data, offset,
size);

The statement that "contents is not meaningful in binary format" seems to be
patently false (or rather, assumes that "binary" is going to be used for a
specific purpose different from mine).

This appears to have been broken since forever:

^252b5132c7 (Richard Henderson 1999-05-03 07:29:11 +0000 284)   /* We don't
want to output anything for a section that is neither
^252b5132c7 (Richard Henderson 1999-05-03 07:29:11 +0000 285)      loaded nor
allocated.  The contents of such a section are not
^252b5132c7 (Richard Henderson 1999-05-03 07:29:11 +0000 286)      meaningful
in the binary format.  */
^252b5132c7 (Richard Henderson 1999-05-03 07:29:11 +0000 287)   if ((sec->flags
& (SEC_LOAD | SEC_ALLOC)) == 0)
b34976b65ae (Alan Modra        2002-11-30 08:39:46 +0000 288)     return TRUE;

-- 
You are receiving this mail because:
You are on the CC list for the bug.