Re: [Oorexx-devel] crash in deadpool

[Rick M. <obj...@gm...>]

I'm able to recreate this crash on Windows now.  It looks like a memory
overlay is occurring.  Interestingly, if I run outside the debugger, it
crashes at the same point it does for you.  If I starting running inside th=
e
debugger initially, it traps at a different location.  This will probably
take me a little while to get the memory overlay bracketed enough that I ca=
n
figure out how to set a watch point for the memory overlay location.
Unfortunately, the addresses move around with each run, so I can't just
record the address and start over :-(

On 2/20/07, Ren=E9 Jansen <rvj...@xs...> wrote:
>
> Hi Rick,
>
> ok, I prepared the method like this:
>
>     inline DeadObject *findFit(size_t length, size_t *realLength)
>      /
> ************************************************************************
> ******/
>      /* Function:  Find first object large enough to satisfy this
> request.  If the */
>      /* pool is empty, this returns NULL.  If a block is returned it
> is removed    */
>      /* from the pool before
> return.                                               */
>      /
> ************************************************************************
> ******/
>      {
>          DeadObject *newObject =3D anchor.next;
>          size_t newLength;
>          int probes =3D 1;
>         printf("1");
>          for (newLength =3D newObject->size(); newLength !=3D 0;
> newLength =3D newObject->size()) {
>         printf("2");
>              if (newLength >=3D length) {
>               printf("3");
>                  if (probes > ReorderThreshold)
>                  {
>                      // we had to examine a lot of objects to get a
> match.
>                      // it's worthwhile percolating the larger
> objects on the rest of the
>                      // chain toward the front.  We only do this when
> we're starting to have problems
>                      // allocating objects because of fragmentation.
>                      DeadObject *tailObject =3D newObject->next;
>                      for (size_t tailLength =3D tailObject->size();
> tailLength !=3D 0; tailLength =3D tailObject->size())
>                      {
>                       printf("4");
>                          // the size we just had problems with is a
> good marker for
>                          // selecting candidates to move toward the
> front.  The will guarantee
>                          // that a similar request for the same size
> will succeed faster in the future.
>                          DeadObject *nextObject =3D tailObject->next;
>                          if (tailLength > length)
>                          {
>                           printf("5");
>                              tailObject->remove();
>                              addSingle(tailObject);
>                          }
>                          tailObject =3D nextObject;
>                         printf("6");
>                      }
>                  }
>                  newObject->remove();
>                 printf("7");
>                  logHit();
>                 printf("8");
>                  *realLength =3D newLength;
>                 printf("9");
>                  return newObject;
>              }
>              probes++;
>              newObject =3D newObject->next;
>          }
>          logMiss();
>          return NULL;
>
>
> and after this sequence of calls the segfault occurs:
>
> [...]
> 789123789123789123789123789123789123789123789123789123789122222222222223
> 789123789122222222222222222222378912378912378912378912378912378912378912
> 378912378912378912378912378912378912378912378912378912378912378912378912
> 378912378912378912378912378912378912378912378912378912378912378912378912
> 378912378912378912378912378912378912378912378912378912378912378912378912
> 222222222222222222237891222222222222222222222378912378912378912378912378
> 912378912378912378912378912378912378912378912378912378912378912378912378
> 912378912378912378912378912378912378912378912378912378912378912378912378
> 912378912378912378912378912378912378912378912378912378912378912378912378
> 912378912378912378912378912378912378912378912378912378912223789123789122
> 237891237891237891237891237891237891237891237891237891237891237891237891
> 237891237891237891237891237891237891237891237891237891237891237891237891
> 237891237891237891237891237891237891237891237891237891237891237891237891
> 237891237891237891237891237891237891237891237891237891237891237891237891
> 237891237891237891237891237891237891237891237891237891237891237891237891
> 237891237891237891237891237891237891237891237891237891237891237891237891
> 237891223789122222222222222222222223789123789122378912378912222222222222
> 222222222237891222222222222222222222222222222222222222222222222222222222
> 222222222222222222222222222222222222222222222222222223464646464564645646
> 464646464646464646464646464646464646456464564646464646464645646464645646
> 464646464564646464646464646464564646456464645645645646464646464564564564
> 645646464646464646464646464646464646464646464646464646464646789122237891
> 22223Segmentation fault
>
>
> Ren=E9.
>
>
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share
> your
> opinions on IT & business topics through brief surveys-and earn cash
> http://www.techsay.com/default.php?page=3Djoin.php&p=3Dsourceforge&CID=3D=
DEVDEV
> _______________________________________________
> Oorexx-devel mailing list
> Oor...@li...
> https://lists.sourceforge.net/lists/listinfo/oorexx-devel
>