Home Original page

[PATCH v3] elf: Keep unused section symbols for Solaris binary

H.J. Lu hjl.tools@gmail.com
Wed Dec 10 02:20:35 GMT 2025 
On Wed, Dec 10, 2025 at 2:12 AM Ali Bahrami <Ali.Bahrami@oracle.com> wrote:
>
> On 12/9/25 00:31, Jan Beulich wrote:
> >> 1. Update bfd_keep_unused_section_symbols to return true for Solaris
> >> binary.
> >> 2. Keep section symbols if input section symbols should be kept and
> >> output target is the same as input target.
> > This is okay afaic, but please give Rainer and/or Ali a chance to comment.
> >
> > Irrespective I have a question here, though (perhaps also to Alan and Nick):
> > Isn't it wrong for objcopy to remove_any_ symbols (including section ones)
> > without an explicit request to do so? IOW shouldn't objcopy without any
> > specific options largely generate an output identical to the input?
>
>
>     I asked the same question fairly early in the bug. I was
> surprised to see a change like this done by default, and also
> that it's an x86-only optimization not done consistently
> across platforms.
>
> I'm sure that the counter argument to all of that is
> that if it's not enabled by default, it won't get used.
>
> Since I'm opining, I'll also ask why this isn't more
> appropriately done only by the link-editor. I'd say
> that if gld doesn't want section symbols, it doesn't
> need to make them, but that objcopy ought (by default)
> to preserve the choice made by the link-editor.

It is

Solaris:

  [17] .bss              NOBITS          00000000001008f8 0008f8
000004 00  WA  0   0  4
  [18] .symtab           SYMTAB          0000000000000000 0008f8
000378 18     19  28  8
  [19] .strtab           STRTAB          0000000000000000 000c70
00007d 00   S  0   0  1
  [20] .SUNW_symtabsort  SUNW_symsort    0000000000000000 000cf0
000018 04     18   0  8
  [21] .SUNW_symtabnsort SUNW_symnsort   0000000000000000 000d08
000028 04     18   0  8
  [22] .comment          PROGBITS        0000000000000000 000d30
000013 01      0   0  1
  [23] .shstrtab         STRTAB          0000000000000000 000d43
0000ea 00   S  0   0  1

vs GNU

  [17] .bss              NOBITS          00000000001008f8 0008f8
000004 00  WA  0   0  4
  [18] .SUNW_symtabsort  SUNW_symsort    0000000000000000 0008f8
000018 04     21   0  8
  [19] .SUNW_symtabnsort SUNW_symnsort   0000000000000000 000910
000028 04     21   0  8
  [20] .comment          PROGBITS        0000000000000000 000938
000013 01      0   0  1
  [21] .symtab           SYMTAB          0000000000000000 000950
000378 18     22  28  8
  [22] .strtab           STRTAB          0000000000000000 000cc8
00007d 00      0   0  1
  [23] .shstrtab         STRTAB          0000000000000000 000d45
0000ea 00      0   0  1

.symtab, .strtab and .shstrtab aren't treated as opaque sections.
Their contents are decoded and stored internally.   During output,
they are written out from the internal presentation in the GNU order.
which is different from Solaris order.

> Anyway, I just tried this latest v3 patch. I had to
> drop the testsuite part, and for reasons I don't
> fully understand, had to recreate the objcopy.c part
> of it manually. That's probably not an upstream issue.
> Having done that, it builds, and it seems to preserve
> the section symbols, so it addresses my main concern.
>
> Thanks.
>
> - Ali
>


-- 
H.J.
More information about the Binutils mailing list