Re: [Valgrind-users] New mem leak detection algorithm finds more bugs

[Nicholas N. <nj...@ca...>]

On Thu, 30 Oct 2003 sf...@ge... wrote:

> I saw the code that skips the stack during leak detection.
>
> I've coded and am currently testing a modification to the memory leak
> detection to flag lost blocks that are pointed to by unused memory
> (free mem or other lost blocks).  E.g. a circular linked list or a
> double linked list will be detected.  As will a leaked block that is
> pointed to by a freed block (it's not enough to free the head of a list,
> you have to walk the list).
>
> Hopefully I'll make Nicholas happy (as per his last line in
> http://sourceforge.net/mailarchive/message.php?msg_id=4892174 ).
>
> I will outline the algorithm below.  Please let me know if you see

Have you considered more standard garbage collection algorithms, eg.
tracing through all the reachable objects from a root set?  I think the
algorithm you described might, in effect, do the same thing, but I wonder
if doing it in a more straightforward GC-style way would be simpler?

> But maybe that is ok.  Good programmers free all their memory
> before program exit.

Really?  Hmm.  Good programmers don't create space leaks either :)

N