[ODE] ODE debugging dSolveLCP

[Erkin Bahceci]

Hi,
We also had some errors regarding dSolveLCP, which were (as we later found out)
caused by high stack size requirements of dWorldStep. The application crashed
when number of collisions reached a certain number. Using dWorldQuickStep instead
fixed the problem in our case. Or you can alternatively increase stack size, but the
same problem may arise at some higher-collision situation again.



On 08 Feb 2005 19:23 EET you wrote:

> Hi.
> 
> I'm a new subscriber and I have some problems that I hope somebody could 
> solve.
> 
> I was doing some memory debugging with valgrind on my program, and I've 
> got a strange warning / error coming from a ODE library.
> 
> Could somebody please explain me if it's a really dangerous problem? It 
> seems to come from the dSolveLCP function.
> I attach the lines I received from valgrind and GDB.
> 
> Thanks
> M.
> 
> 
> Valgrind Output during normal execution (no segmentation fault yet):
> 
> ==21128== Use of uninitialised value of size 8
> ==21128==    at 0x80AA15D: swapProblem(double**, double*, double*, 
> double*, double*, double*, int*, int*, int*, int, int, int, int, int) 
> (lcp.cpp:240)
> ==21128==    by 0x80AAA6C: dLCP::dLCP(int, int, double*, double*, 
> double*, double*, double*, double*, double*, double*, double*, double*, 
> double*, int*, int*, int*, int*, double**) (lcp.cpp:792)
> ==21128==    by 0x80AC3CF: dSolveLCP(int, double*, double*, double*, 
> double*, int, double*, double*, int*) (lcp.cpp:1137)
> ==21128==    by 0x80A5CD8: dInternalStepIsland_x2(dxWorld*, dxBody* 
> const*, int, dxJoint* const*, int, double) (step.cpp:835)
> 
> The problem recognised by valgrind is then passed to gdb, that gives the 
> following lines:
> 
> 0x080aa15d in swapProblem (A=0x52be9830, x=0x52bf14e0, b=0x52bf16d0, 
> w=0x52bf12f0, lo=0x52bfc800, hi=0x52bfc610, p=0x52be9730, state=0x52be9530,
>      findex=0x52bfc514, n=59, i1=5, i2=6, nskip=60, do_fast_row_swaps=0) 
> at ode/src/lcp.cpp:240
> 240       tmp = w[i1];
> 
> 
> 
> 
> 
> When a Segmentation Fault occurs in my program, most of the time this is 
> the gdb output:
> 
> #0  0x080ac35f in dSolveLCP(int, double*, double*, double*, double*, 
> int, double*, double*, int*) (n=353, A=0xbfeeced0, x=0xbfeeae00, 
> b=0xbfeeb920,
>      w=0xbfeea2e0, nub=15, lo=0xbfff97d0, hi=0xbfff8cb0, 
> findex=0xbfff8714) at ode/src/lcp.cpp:1137
> 1137      dLCP lcp 
> (n,nub,A,x,b,w,lo,hi,L,d,Dell,ell,delta_w,state,findex,p,C,Arows);
> 
> #1  0x080a5cf1 in dInternalStepIsland_x2(dxWorld*, dxBody* const*, int, 
> dxJoint* const*, int, double) (world=0x8146890, body=0xbffff030, nb=42,
>      _joint=0xbfffebd0, nj=85, stepsize=0.0025000000000000001) at 
> ode/src/step.cpp:835
> 835         dSolveLCP (m,A,lambda,rhs,residual,nub,lo,hi,findex);
> 
> 
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>