[PATCH] ld: testsuite: xfail vers26b3 on Solaris/amd64

[Jan Beulich]

On 18.08.2025 16:16, Rainer Orth wrote:
> Hi Jan,
> 
>> On 15.08.2025 10:28, Rainer Orth wrote:
>>> After enabling the ld-elfvers tests on Solaris/x86, one of them FAILs on
>>> Solaris/amd64 only:
>>>
>>> FAIL: vers26b3
>>>
>>> /ld-new: tmpdir/vers26b3.o: relocation R_X86_64_32 against symbol `foo' can not be used when making a shared object; recompile with -fPIC
>>> ./ld-new: failed to set dynamic section sizes: bad value
>>>
>>> The error is wrong: vers26b3.o *was* compiled with -fPIC and repeating
>>> the link with /bin/ld just works.
>>
>> The error, aiui, is issued on the assumption that a compiler wouldn't emit
>> this kind of relocation when building PIC code. I came across that before
>> and found it questionable; I expect H.J. would be able to shed further
>> light on this.
>>
>>> Also, the test is already xfail'ed for Linux/x86_64, so this patch does
>>> the same on Solaris/amd64.
>>
>> Albeit the reason for the xfail on Linux is different, as per the comment
>> there.
>>
>>> Tested on {amd64,i386}-pc-solaris2.11 and {x86_64,i686}-pc-linux-gnu.
>>>
>>> Ok for trunk?
>>
>> Why is it that you add a two part conditional:
>>
>> } elseif { [istarget *86*-*-solaris2*] && [is_elf64 $tmpdir/vers26b2.o] } then {
>>
>> Wouldn't
>>
>> } elseif { [istarget x86_64-*-solaris2*] } then {
>>
>> do? At which point this could perhaps be combined with the existing
>> conditional.
> 
> the idea here is that this would happen both for an amd64-pc-solaris2.11
> gcc and an i386-pc-solaris2.11 one with -m64 (and gcc on Solaris is
> built with multilib support by default): the target triple only reflects
> the default ABI IIUC.  I vaguely remember that there where issues with
> testing binutils with both -m32 and -m64, so I don't usually test that
> case but have both 32-bit-default and 64-bit-default gcc's in parallel.

But are these tests actively being run as 64-bit ones when the primary
target is ix86? (I expect not.)

Jan