Re: [Valgrind-users] 0 bytes definitely lost

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

On Fri, 10 Oct 2003, Russ Fink wrote:

> I'm a semi-newbie to valgrind, and a positive newbie to the application I'm
> debugging.  I'm getting a message from Valgrind that I don't understand; can
> you help me?
>
> I'm running with --leak-check=yes, --show-reachable=yes.  Valgrind is
> saying:
>
> ==5843== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)
> ==5843== malloc/free: in use at exit: 3036062 bytes in 3230 blocks.
> ==5843== malloc/free: 24202 allocs, 20972 frees, 9377725 bytes allocated.
> ==5843== For counts of detected errors, rerun with: -v
> ==5843== searching for pointers to 3230 not-freed blocks.
> ==5843== checked 6748468 bytes.
> ==5843==
> ==5843== 0 bytes in 383 blocks are definitely lost in loss record 1 of 9
> ==5843==    at 0x4002B905: malloc (vg_replace_malloc.c:153)
> ==5843==    by 0x80507F8: ...
>
> What does it mean, "0 bytes in 383 blocks are definitely lost?"
> Semantically, if I lose zero bytes, then I'm not leaking.  I don't believe
> my logic, though - I believe a leak is taking place, but for some reason,
> Valgrind is unable to quantify the amount being leaked.

Could you be leaking zero-sized heap blocks?  If you do leak a zero-sized
block, you really are losing memory because every heap block has
administration bytes at the start and end.

N