Mark,

This should be fixed in upcoming 1.8.6 release (October 15). Hopefully a 
pre-released tar ball will be available for testing next week. We will announce 
it on FORUM.

Elena
On Sep 18, 2010, at 5:09 PM, Mark Miller wrote:

> FYI, I've been seeing same memcpy src/dst overlap problem on our BG/P
> system. It seems to be happening in one of the first attributes ever
> written to the file, a tiny single 4 byte integer value'd attribute. The
> pointer for both src and dst is the same value.
> 
> I cannot find a way to re-produce on something 'small'. 
> 
> Mark
> 
> On Tue, 2010-08-24 at 06:13, Quincey Koziol wrote:
>> Hi marc,
>> 
>> On Aug 24, 2010, at 7:32 AM, Marc POINOT wrote:
>> 
>>> Quincey Koziol wrote:
>>>>    Your code looks reasonable.  We've been working on valgrind issues 
>>>> recently - can you check out the latest code from:
>>>>    http://*svn.hdfgroup.uiuc.edu/hdf5/branches/hdf5_1_8/
>>>>    See if these recent changes help.
>>> 
>>> Great, this removes all Valgrind weird messages I had on some hdf5 calls.
>> 
>>      Good!  :-)
>> 
>>> There's still one I'm investigating now (see below), only one test in my 
>>> test suite
>>> raises this message, I'm trying to find out what's different in this test.
>>> 
>>> ==24685== Syscall param write(buf) points to uninitialised byte(s)
>>> ==24685==    at 0x3E5070B012: __write_nocancel (in 
>>> /lib64/tls/libpthread-2.3.4.so)
>>> ==24685==    by 0x4C589C4: H5FD_sec2_write (H5FDsec2.c:839)
>>> ==24685==    by 0x4C4E603: H5FD_write (H5FDint.c:184)
>>> ==24685==    by 0x4C32FC4: H5F_accum_write (H5Faccum.c:580)
>>> ==24685==    by 0x4C34F7F: H5F_block_write (H5Fio.c:162)
>>> ==24685==    by 0x4D2158F: H5O_flush (H5Ocache.c:486)
>>> ==24685==    by 0x4BD765D: H5C_flush_single_entry (H5C.c:7606)
>>> ==24685==    by 0x4BC9F84: H5C_flush_cache (H5C.c:1801)
>>> ==24685==    by 0x4BA1EB6: H5AC_flush (H5AC.c:843)
>>> ==24685==    by 0x4C2CA29: H5F_flush (H5F.c:1685)
>>> ==24685==    by 0x4C2A044: H5F_dest (H5F.c:994)
>>> ==24685==    by 0x4C2D2F3: H5F_try_close (H5F.c:1909)
>>> ==24685==  Address 0x534713C is 6,852 bytes inside a block of size 8,192 
>>> alloc'd
>>> ==24685==    at 0x4905E12: realloc (vg_replace_malloc.c:306)
>>> ==24685==    by 0x4CFCA97: H5MM_realloc (H5MM.c:140)
>>> ==24685==    by 0x4C320C2: H5F_accum_adjust (H5Faccum.c:335)
>>> ==24685==    by 0x4C32558: H5F_accum_write (H5Faccum.c:416)
>>> ==24685==    by 0x4C34F7F: H5F_block_write (H5Fio.c:162)
>>> ==24685==    by 0x4D2158F: H5O_flush (H5Ocache.c:486)
>>> ==24685==    by 0x4BD765D: H5C_flush_single_entry (H5C.c:7606)
>>> ==24685==    by 0x4BCA036: H5C_flush_cache (H5C.c:1824)
>>> ==24685==    by 0x4BA1EB6: H5AC_flush (H5AC.c:843)
>>> ==24685==    by 0x4C2CA29: H5F_flush (H5F.c:1685)
>>> ==24685==    by 0x4C2A044: H5F_dest (H5F.c:994)
>>> ==24685==    by 0x4C2D2F3: H5F_try_close (H5F.c:1909)
>> 
>>      Well, I didn't say we were completely done yet.  ;-)  If you can get 
>> this warning (which is most likely harmless) down to a simple program, I can 
>> see about fixing it.
>> 
>>      Quincey
>> 
>> 
>> _______________________________________________
>> Hdf-forum is for HDF software users discussion.
>> [email protected]
>> http://*mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
> -- 
> Mark C. Miller, Lawrence Livermore National Laboratory
> ================!!LLNL BUSINESS ONLY!!================
> [email protected]      urgent: [email protected]
> T:8-6 (925)-423-5901    M/W/Th:7-12,2-7 (530)-753-8511
> 
> 
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]
> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org


_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org

Reply via email to