Hi Dietmar
I learned (sometimes very hard) that jumping to conclusion state without checking the technical arguments from the other side will not bring business success. Let me question your statement " the answer to the segfaults is given".
What I see is that:
- the compiler is generating a instruction that we question whether it's supported by Quark or not
If the instruction is not supported by Quark, we should see a 'invalid instruction exception', not a segfault.
The question, that Aurelien (in the bug report) has, is to investigate if the lock instruction is resulting in the correct output.
If I find some day later, I'll check in the instruction reference and read about the 'lock' instruction. Maybe I write a small asm program to see if the instruction behaves correctly. (But time on my side is limited).
Until we have the result, may I ask to hold the horses and spread the message "the answer is given". 
Tom