Re: [PATCH hail] chunkd: don't leak an FS object iterator

[Jeff Garzik <jeff@garzik.org>]

On 09/29/2010 11:20 AM, Jim Meyering wrote:
>
> chk_list_objs called fs_list_objs_open without also calling
> fs_list_objs_close.
>
>   32,808 bytes in 1 blocks are definitely lost in loss record 413 of 419
>      at 0x4A0515D: malloc (vg_replace_malloc.c:195)
>      by 0x31BA8A26D0: __alloc_dir (opendir.c:184)
>      by 0x405619: fs_list_objs_open (be-fs.c:974)
>      by 0x40B202: chk_list_objs (selfcheck.c:41)
>      by 0x40B575: chk_dbscan (selfcheck.c:131)
>      by 0x40B628: chk_thread_scan (selfcheck.c:147)
>      by 0x40B757: chk_thread_command (selfcheck.c:179)
>      by 0x40B890: chk_thread_func (selfcheck.c:219)
>      by 0x31BC464E83: g_thread_create_proxy (gthread.c:1893)
>      by 0x31BB407760: start_thread (pthread_create.c:301)
>      by 0x31BA8E151C: clone (clone.S:115)

After seeing a few valgrind references from you, I'm curious... do you 
by chance happen to have a valgrind suppression file for openssl on Fedora?

I've been wanting to run valgrind on chunkd, but each time I attempt it, 
I -- and valgrind -- have been overwhelmed by openssl false positives. 
openssl, deep in its RAND_xxx functions, intentionally does crazy stuff 
like using random, uninitialized stack contents as RNG entropy.  Cute, 
but valgrind quite rightly complains loudly about it.

It's a topic I've been meaning to research, because I currently lack the 
valgrind-fu necessary to have an effective valgrind+chunkd session.

Thanks,

	Jeff