Re: [Valgrind-developers] valgrind: r12586: Fix false positive in sys_clone on amd64 when optional args are not given (e.g. child_tidptr)

[Christian B. <bor...@de...>]

>> I remember that Christian looked at that testcase failure and r12033
>> fixed it. I have little wisdom in these syscalls, so I'm copying
>> Christian. Would be good to get this nailed.
> Yes, r12033 introduced the fix for the *SETTID cases.
>>From what I understand, a similar problem can also happen with SETTLS,
> which I believe is not fixed in s390x.
> 
> So, the summary on linux is:
> x86 and amd64 have *SETTID and SETTLS ok.
> s390x has *SETTID ok.

Seems that s390 doesnt look at the TLS value at all for definedness checking.
So we dont have a false positive here, instead we might miss a real error.
Will fix.

Christian