Re: valgrind - gw_check_leaks() - memory leaks etc

        Hi,

On Mon, Jan 23, 2006, fred wrote:
>     If someone could take a look at some of these....
>     Eg invalid read size 4 in lock....ie
>     static void lock(List *list)
>     {
>         gw_assert(list != NULL);
>         mutex_lock(list->single_operation_lock);
>     }
>     what is wrong here? anyone? and why would valgrind report this...kannel
>     seems to be working fine........

 Some memory allocations are often zeroed out or have a predictable
 content, it could explain why kannel works despite accessing
 uninitialized memory.

>     for the Question here I'm more concerned about those invalid reads...to do
>     with the library code....

 I don't understand your question.

>     ==3368== Thread 6:
>     ==3368== Invalid read of size 4
>     ==3368==    at 0x80D2683: lock (list.c:525)
>     ==3368==    by 0x80D2265: gwlist_remove_producer (list.c:398)
>     ==3368==    by 0x8057EDA: smsboxc_run (bb_boxc.c:1196)
>     ==3368==    by 0x80C903F: new_thread (gwthread-pthread.c:346)
>     ==3368==    by 0x40B6DFB: start_thread (in /lib/tls/libpthread-0.60.so)
>     ==3368==    by 0x438DF29: clone (in /lib/tls/libc-2.3.2.so)

 If this is the read you mention, then you should get a fresher kannel
 build or remove your local changes since there's no
 gwlist_remove_producer call in smsboxc_run at bb_boxc.c:1196 (in CVS).

   Cheers,

-- 
Loïc Minier <[email protected]>
Current Earth status:   NOT DESTROYED