On 08-15 09:39, Sergei Golovan wrote: > Hi Witold! > > On Fri, Aug 13, 2010 at 5:53 AM, Witold Baryluk > <bary...@smp.if.uj.edu.pl> wrote: > > > > Simple test case: > > > > $ cd /usr/lib/erlang/lib/wx-0.98.6/examples/sudoku > > $ erl -smp true > > 1> sudoku:start(). > > > > > > wait few seconds, eventually interact with widget. > > > > For few seconds it will make CPU 100% busy (probably bug in timiing or > > epoll) > > > > After a while it will crash whole VM with "Segmentation fault". > > Finally, I've managed to get a segfault when running erl -smp +S 1 > (SMP with one core, like in your case). If the number of cores is > greater than one Erlang and wx work fine. > > Could you try to run erl -smp +S 2 on your one-core processor to check > if this can be used as a workaround?
Thanks for quick answer. erl -smp true +S 2 on my one-core preocessor, still make seg.fault. What is interesting, many other wx demos (and my application which i write in Erlang+wx), seems to be working fine. Only sudoku segfault so quickly. This can be bug in application, becuase i found in wx documentation that there are few places, where memory managment is manual, and one need to be very cerful to not crash VM. But the fact that it works in real smp, means that it is bug or race-condition somehwere (in sudoku or in VM). I have dual-core machine, but it is lenny, so there is no wx installed, and cannot test it now. -- Witold Baryluk
signature.asc
Description: Digital signature
