[PATCH v2] gas: add --reloc-section-sym={all, internal, none} option for ELF

[Jan Beulich]

On 23.02.2026 18:36, Fangrui Song wrote:
> On Sun, Feb 22, 2026 at 11:20 PM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 22.02.2026 03:03, Fangrui Song wrote:
>>> --- a/gas/testsuite/gas/elf/elf.exp
>>> +++ b/gas/testsuite/gas/elf/elf.exp
>>> @@ -4,15 +4,15 @@
>>>  # it under the terms of the GNU General Public License as published by
>>>  # the Free Software Foundation; either version 3 of the License, or
>>>  # (at your option) any later version.
>>> -#
>>> +#
>>>  # This program is distributed in the hope that it will be useful,
>>>  # but WITHOUT ANY WARRANTY; without even the implied warranty of
>>>  # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>>>  # GNU General Public License for more details.
>>> -#
>>> +#
>>>  # You should have received a copy of the GNU General Public License
>>>  # along with this program; if not, write to the Free Software
>>> -# Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA.
>>> +# Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA.
>>>
>>>  #
>>>  # elf tests
>>> @@ -170,30 +170,32 @@ if { [is_elf_format] } then {
>>>      switch -glob $target_triplet {
>>>       alpha*-*-* { }
>>>       am3*-*-* { }
>>> +     avr-*-* { }
>>>       *c54x*-*-* { }
>>>       cr16*-*-* { }
>>>       crx*-*-* { }
>>>       h8300-*-* { }
>>> -     hppa*-*-* { }
>>> -     iq2000*-*-* { }
>>> -     mips*-*-* { }
>>> +     loongarch*-*-* { }
>>> +     m681*-*-* { }
>>> +     m68hc*-*-* { }
>>>       mn10200-*-* { }
>>>       mn10300-*-* { }
>>>       msp43*-*-* { }
>>> -     rl78-*-* { }
>>>       riscv*-*-* { }
>>> +     rl78-*-* { }
>>>       rx-*-* { }
>>> -     loongarch*-*-* { }
>>> +     xgate-*-* { }
>>> +     vax-*-* { }
>>>       default {
>>> -         # The next test can fail if the target does not convert fixups
>>> -         # against ordinary symbols into relocations against section symbols.
>>> -         # This is usually revealed by the error message:
>>> -         #  symbol `sym' required but not present
>>> -         setup_xfail "m681*-*-*" "m68hc*-*-*" "xgate-*-*" "vax-*-*" "avr-*-*"
>>> -         run_dump_test redef
>>> +         # The following tests require that the target adjusts relocations
>>> +         # referencing local binding symbols to be against section symbols.
>>>           run_dump_test equ-reloc
>>> +         run_dump_test redef
>>> +         run_dump_test "reloc-section-sym-all"
>>> +         run_dump_test "reloc-section-sym-internal"
>>>       }
>>>      }
>>> +    run_dump_test "reloc-section-sym-none"
>>>      run_dump_test "pseudo"
>>>      run_dump_test "text-prev" $dump_opts
>>>      run_dump_test "text-subsect" $dump_opts
>>> @@ -245,7 +247,7 @@ if { [is_elf_format] } then {
>>>      # in the symbol type test.
>>>      # We also need to exclude targets that do not support unique objects.
>>>      if {    [istarget "*-*-hpux*"]
>>> -      || [istarget "arm*-*-*"]
>>> +      || [istarget "arm*-*-*"]
>>>           || [istarget "mips*-*-*"]
>>>           || [istarget "msp*-*-*"]
>>>        || [istarget "visium-*-*"]
>>> @@ -380,7 +382,7 @@ if { [is_elf_format] } then {
>>>      run_dump_test "startof"
>>>
>>>      run_dump_test "missing-build-notes"
>>> -
>>> +
>>>      run_dump_test "common1"
>>>      run_dump_test "common2"
>>>
>>
>> This became much more of a change than what v1 had, and I question parts thereof.
>> You're dropping XFAILs. If there's a proper reason for doing so, it needs stating
>> in the description. You're further excluding the equ-reloc test for more targets.
>> This pretty likely is wrong?
>>
>> You apparently also try to re-sort the list of excluded targets in that switch,
>> yet then vax ends up after xgate.
> 
> I have used a build-all-gas.sh script to build 40+ triplets, including
> every triplet mentioned in elf.exp.
> I have determined that the equ-reloc and redef tests actually pass on
> the disabled hppa and iq2000.

That's good, but definitely wants to be a separate change then, imo.

> The list is now sorted alphabetically.

Only quite, as mentioned.

> vax is special: its tc_fix_adjustable macro in gas/config/tc-vax.h
> includes the condition S_IS_LOCAL ((FIX)->fx_addsy)), causing
> reloc-section-sym-all to fail.
> setup_xfail only works for the next run_dump_test. I think it's easier
> to just exclude it.
> 
> I feel like xfail is often just a placeholder for 'this target is
> weird but I don't have time to fix it.'

But isn't this what xfail is for (among other purposes)? When you make
changes to common code and don't know a particular target (hence also
not why it may exhibit "weird" behavior), marking tests this way is
often the best you can do. Diligent arch maintainers would then pick
this up, converting to an exclusion or (preferably) addressing the
issue.

>> I'm also unconvinced all the whitespace changes should go here.
> 
> While I understand the issue of interleaving functional changes with
> whitespace noise, I noticed our .editorconfig has
> trim_trailing_whitespace = true enabled, which is causing neovim to
> automatically scrub spaces on save.
> To save everyone the headache of messy diffs, we should probably do a
> one-time sweep in batch.

Possibly, albeit doing so also has a perhaps significant downside: It
can badly get in the way of "git blame" or alike.

>> With changes like these, btw, a revlog would also have been helpful.
> 
> What is revlog?

A brief indication of what changed compared to earlier versions of a
patch, outside of the commit message area.

Jan