Re: Segfaults in lj_snap_shrink

  • From: Denis Golovan <denis.golovan@xxxxxxxxx>
  • To: luajit@xxxxxxxxxxxxx
  • Date: Wed, 30 Nov 2016 12:54:37 +0200

I use the following options:
jit.opt.start("maxmcode=10240", "maxtrace=3000", "maxside=400",
"maxsnap=3000", "maxirconst=2000",
                 "hotloop=5", "hotexit=20", "instunroll=25",
"loopunroll=100", "callunroll=25")

Thanks, I'll try your suggestion and come back.

2016-11-30 12:34 GMT+02:00 Vyacheslav Egorov <dmarc-noreply@xxxxxxxxxxxxx>:

Are you running with default JIT parameters - or did you tweak any?

I have seen this sort of thing happen when things overflow uint16_t.

You can try running with:

if (((MSize)(uint16_t)(snap->mapofs + m)) != (snap->mapofs + m)) abort();

to check if this is the case.


Other related posts: