Hi all,
Maybe this is an oft repeated query, but I would still appreciate any
feedback.
I am trying to find out a memory leak in my program, and while looking
at the Valgrind report I see a lot of these statements.
36 bytes in 4 blocks are definitely lost in loss record 1,865 of 3,671
==18557== at 0x4C27FA4: malloc (vg_replace_malloc.c:296)
==18557== by 0x56E43B0: H5MM_xstrdup (in
/*/common/soft/hdf5/1.8.12/@sys/intel/14.0/impi-4.1.3/lib/libhdf5.so.8.0.1)
==18557== by 0x5805138: H5Tget_member_name (in
/*/common/soft/hdf5/1.8.12/@sys/intel/14.0/impi-4.1.3/lib/libhdf5.so.8.
0.1)
==18557== by 0x4501E2: HDF5HeaderDataattributeScan (in
/*/EXEC_6.3.0_v26_odin_PSAMR)
==18557== by 0x55D211F: H5A_attr_iterate_table (in
/*/common/soft/hdf5/1.8.12/@sys/intel/14.0/impi-4.1.3/lib/libhdf5.s
o.8.0.1)
==18557== by 0x56F59E3: H5O_attr_iterate_real (in
/*/common/soft/hdf5/1.8.12/@sys/intel/14.0/impi-4.1.3/lib/libhdf5.so
.8.0.1)
==18557== by 0x56F6F23: H5O_attr_iterate (in
/*/common/soft/hdf5/1.8.12/@sys/intel/14.0/impi-4.1.3/lib/libhdf5.so.8.0.
1)
==18557== by 0x55C91A1: H5Aiterate2 (in
/*/common/soft/hdf5/1.8.12/@sys/intel/14.0/impi-4.1.3/lib/libhdf5.so.8.0.1)
I looked at some queries relating to this, where it's said that there's
a patch for this in version 1.8. I am using HDF5 library version 1.8.12.
https://lists.web.psi.ch/pipermail/h5part/2012-May/000207.html
Also the link below mentions this issue while writing out in parallel mode.
http://www.hdfgroup.org/HDF5/faq/valgrind.html
My query is that do I need to have the HDF5 library compiled in a
certain way, the libhdf5.settings file show that the use of memchecker
is switched off, or can I ignore these messages from Valgrind. I worry
that this makes the code rather hard to port, as I need to have the HDF5
library compiled with the "memchecker" option switched on.
I will be glad to know any suggestions/work arounds etc.
Nitya