Re: SIGSEGV in LuaJIT 2.1 VM
- From: Tomas Kvasnicka <nzt4567@xxxxxxx>
- To: luajit@xxxxxxxxxxxxx
- Date: Fri, 18 Nov 2016 16:36:06 +0100
This isn't overly surprising - the faulting instruction is the one
which sets the current-trace variable, so until it executes, lvmst
will be reporting the old value of the variable.
Thanks for the explanation. I was suspecting the core files to be maybe
corrupted - I am glad that’s not the case.
For the sake of investigation, you say "mostly", so there are different chips
as well. Could you possibly list models where the issue occurs and where it
does not occur for you? If it's not a sensitive info, ofc.
Debugging always comes first! :) So far I can say that it definitely happens
with E5-1620 v2, E5-2620 v2, E5-2620 v3 and E5-2620 v4. There are also chips
such as E3-1231 or E3-1230 and some others as well but I can’t reliably say
anything about these as we have not been collecting this info until now -
hardware issue was not even close to what we were suspecting. I can start
storing this info now and update you next week - I am wondering if there is at
least one chip where this doesn’t happen at all.
Thank you guys!
Other related posts:
