Re: [F4-users] opcode debugging on Valgrind..

[Charles J. L. Q. C. <cha...@ho...>]

  Valgrind is a memory debugger for *nix,

  note that opcode _is_ working now (i.e: it detects collisions) but, under 
certain circunstances which im trying to find, it stops detecting some of 
the faces,

Ok, i narrowed the problem where valgrind is reporting memory errors to this 
area:
(just look the file lines on the report):

  AABBTreeCollider::InitQuery

// Setup matrices
Matrix4x4 InvWorld0, InvWorld1;
if(world0) InvertPRMatrix(InvWorld0, *world0); <--- here occurs the
else InvWorld0.Identity();

if(world1) InvertPRMatrix(InvWorld1, *world1); <--- first errors
else InvWorld1.Identity();

Matrix4x4 World0to1 = world0 ? (*world0 * InvWorld1) : InvWorld1; <-- here 
occurs
Matrix4x4 World1to0 = world1 ? (*world1 * InvWorld0) : InvWorld0; <-- after



inside InvertPRMatrix, the buggy write seems to originate here:

dest.m[3][2] = -(src.m[3][0]*src.m[2][0] + src.m[3][1]*src.m[2][1] + 
src.m[3][2]*src.m[2][2]);


Could be Matrix4x4::m is messed up?  but its a float m[4][4], shouldnt be

>From: "Pierre Terdiman" <p.t...@wa...>
>Reply-To: fla...@li...
>To: <fla...@li...>
>Subject: Re: [F4-users] opcode debugging on Valgrind..
>Date: Thu, 29 Aug 2002 17:33:16 +0200
>
>(What is Valgrind ?)
>
> >   Use of uninitialised value of size 4
> > ==29438==    at 0x438843CA: IceMaths::InvertPRMatrix(IceMaths::Matrix4x4
>&,
> > IceMaths::Matrix4x4 const &) 
>(plugins/collide/opcode/OPC_Matrix4x4.cpp:69)
> >
> > ....
> >
> >    Use of uninitialised value of size 4
> > ==29438==    at 0x43897C5B: ???
>(plugins/collide/opcode/OPC_Matrix4x4.h:297)
> > ==29438==    by 0x43891D12:
> > Opcode::AABBTreeCollider::InitQuery(IceMaths::Matrix4x4 const *,
> > IceMaths::Matrix4x4 const *)
> > (plugins/collide/opcode/OPC_TreeCollider.cpp:265)
> >
> >
> > and at destruction i get also some of these:
> >
> > Mismatched free() / delete / delete []
> > ==29438==    at 0x40046819: __builtin_vec_delete (vg_clientfuncs.c:201)
> > ==29438==
> >
> >
> >   What do you think Pierre?
>
>Errr, change your compiler ?
>
>I don't know what to say without more infos, except I doubt there are some
>bugs here. A delete / delete [] mismatch is still possible anyway. Do you
>know where it's supposed to happen, exactly ?
>
>Pierre
>
>
>
>-------------------------------------------------------
>This sf.net email is sponsored by:ThinkGeek
>Welcome to geek heaven.
>http://thinkgeek.com/sf
>_______________________________________________
>flatfour-users mailing list
>fla...@li...
>https://lists.sourceforge.net/lists/listinfo/flatfour-users




_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com