[PATCH v2 1/2] binutils, gas, ld: Disable SFrames for some tests

[Jan Beulich]

On 25.09.2026 12:14, Sam James wrote:
> Jan Beulich <jbeulich@suse.com> writes:
> 
>> On 25.09.2026 09:30, Sam James wrote:
>>> --- a/ld/testsuite/ld-x86-64/x86-64.exp
>>> +++ b/ld/testsuite/ld-x86-64/x86-64.exp
>>> @@ -30,6 +30,10 @@ set emul "elf_x86_64"
>>>  set saved_ASFLAGS "$ASFLAGS"
>>>  set ASFLAGS "$ASFLAGS -mx86-used-note=no"
>>>  
>>> +if { [gas_sframe_check] } then {
>>> +    set NO_SFRAME_ASFLAGS "--gsframe=no"
>>> +}
>>
>> With this, ...
>>
>>> @@ -586,12 +590,14 @@ run_dump_test "tls-le-pic-3"
>>>  run_dump_test "tls-le-pic-3-x32"
>>>  
>>>  if { ![skip_sframe_tests] } {
>>> +    set NO_SFRAME_ASFLAGS "--gsframe=no"
>>> +
>>
>> ... why this further (and unconditional) setting?
> 
> Sorry, pasto. I'd changed my mind and didn't clean it up.
> 
> The first one is the one that should stay.

With this adjusted, the change is okay except for ...

>>>      run_dump_test "sframe-simple-1"
>>>      run_dump_test "sframe-link-1"
>>>      # Test --discard-sframe
>>>      run_dump_test "sframe-command-line-1"
>>>      # Test --no-ld-generated-unwind-info
>>> -    run_dump_test "sframe-command-line-2"
>>> +    run_dump_test "sframe-command-line-2" "{as $NO_SFRAME_ASFLAGS}"
>>>      run_dump_test "sframe-reloc-1"
>>>      run_dump_test "sframe-gc-sections-1"
>>>      run_dump_test "sframe-gc-sections-2a"
>>
>> And then, as we're in a section of SFrame specific tests, why would it
>> want disabling here at all?
> 
> The test is checking whether an object gets "upgraded" to SFrames
> unduly. If SFrames are default-enabled, the test fails becuse it thinks
> the upgrade happened. I think ideally would have a way to check if it is
> default-enabled or not and only disable in that case, but it's one test,
> and from my reading of it, I don't think passing --gsframe=no undermines
> the intent of it.

... this one aspect, which I'd like Indu to confirm.

Jan