HDF5 version 1.6.6-pre1 currently under development ================================================================================ INTRODUCTION ============ This document describes the differences between HDF5-1.6.5 and HDF5-1.6.5-*. It contains information on the platforms tested and known problems in HDF5-1.6.5. For more details, check the HISTORY.txt file in the HDF5 source. The HDF5 documentation can be found on the NCSA ftp server (ftp.ncsa.uiuc.edu) in the directory: /HDF/HDF5/docs/ Documentation for the current release is also on the HDF web site: http://hdf.ncsa.uiuc.edu/HDF5/doc/ For more information look at the HDF5 home page at: http://hdf.ncsa.uiuc.edu/HDF5/ If you have any questions or comments, please send them to: hdfhelp@ncsa.uiuc.edu CONTENTS ======== - New Features - Support for New Platforms, Languages and Compilers - Bug Fixes since HDF5-1.6.5 - Documentation - Platforms Tested - Supported Configuration Features Summary - Known Problems New Features ============ Configuration: ------------------------- Source code distribution: ------------------------- Library: -------- Parallel Library: ----------------- Tools: ------ - h52gif and gif2h5: Both these tools were revised to include the High Level Image API support, and tests were added to /hl/tools/gif2h5. PVN - 2007/04/13 - h5dump: added support for double long type H5T_NATIVE_LDOUBLE. PVN 2007/03/13 - h5dump: added support for binary output, see usage. PVN 2007/03/13 - h5repack: added support for reading and writing by hyperslabs for large files. PVN - 2007/03/01 - h5repack: a new option allows the copy using the file type (default) instead of the previous conversion to native type. PVN - 2007/03/01 - h5repack: output the percentage of compression used. PVN - 2007/03/01 - h5diff: added support for comparing dataset regions. PVN - 2007/02/20 - h5diff: added support for reading and comparing by hyperslabs for large files. PVN - 2007/02/20 - h5diff: printing of dataset dimensions along with dataset name. PVN - 2007/02/19 F90 API: -------- C++ API: -------- - Added support for szip: DSetCreatPropList::setSzip - BMR 2007/01/24 Support for New Platforms, Languages and Compilers ================================================== - Added support for MAC Intel (Darwin 8.8.2) (gcc 4.0.1 and g95 0.91) EIP - 2007/02/13 Configuration ------------- Bug Fixes since HDF5-1.6.5 Release ================================== Library ------- - Fixed several bugs with writing fill values for datasets that have a variable-length datatype or component datatype. QAK - 2007/06/19 - STDIO driver didn't support files bigger than 2GB because the OFFSET parameter of fseek is of type LONG INT, not big enough for big files. Use fseeko instead for big files if it's available on the system. SLU - 2007/4/5 - Relaxed restrictions on attribute operations to allow a file ID to be used as the "location ID". If a file ID is used, the attribute operation will occur on the root group of the file. - QAK - 2007/02/09 - Fixed a bug in H5Sselect_valid() that caused an incorrect value to be returned (0) instead of FAIL on error confitions such as the selection or extent not being defined. CMC - 2007/01/05 - Fixed potential file corruption bug which could overwrite a portion of an object's header when an attribute was renamed. If the new name for the attribute was longer than the old name, it's possible that the attribute could grow enough to overwrite another message in the object's header. QAK - 2007/01/02 - Fixed file corruption bug which could write an incorrect number of messages to an object's header under certain circumstances. Generally, the sequence of actions to generate this bug looks like this: - Create an object - Close the file - Re-open the file - Add 2 (or more) attributes to the object - Close the file - Re-open the file - Delete one of the attributes on the object - Add a smaller attribute to the object - Delete the smaller atttribute on the object - Add a larger attribute on the object After this, the number of header messages stored for the object will be off by one. Other sequences of modifying attributes on an object could also trigger this bug. If you are opening an object and the bottom few messages of the HDF5 error stack resembles this, the object has been affected by this bug: #007: ../../hdf5_v1.6/src/H5C.c line 3887 in H5C_load_entry(): unable to load entry major(08): Meta data cache layer minor(40): Unable to load metadata into cache #008: ../../hdf5_v1.6/src/H5Ocache.c line 332 in H5O_load(): corrupt object header - too few messages major(12): Object header layer minor(40): Unable to load metadata into cache Specifically, "corrupt object header" is the best string to search for in the HDF5 error stack output. If your files have been affected by this bug, or you are concerned that your files might have been, please contact the HDF Helpdesk at hdfhelp@ncsa.uiuc.edu for a tool to detect and repair files affected by this problem. QAK - 2006/6/16 - Fixed various problems with retrieving names of objects, especially with mounted files. QAK - 2005/12/25 Configuration ------------- - Changed to default to --disable-shared if parallel is enabled. - AKC 2007/5/12 - Correct an coding error in configure when it tries to locate the needed MPI and MPI-IO library for the fortran interface. - AKC 2007/5/9. - Changed default fortran compiler to g95 when gcc is used. - AKC 2007/2/17. - Configure can now use any tr command. No more need for defining variable TR nor is it supported. -AKC 2006/05/20 Performance ------------- Tools ----- - Fixed a bug on h5dump regarding the display of named datatypes attributes. PVN 2007/03/13 - Fixed a bug on h5dump regarding the display of group comments. PVN 2007/03/13 - Fixed a bug on h5dump regarding the display of hardlinks pointing to the root group. PVN 2007/03/13 - Fixed a bug in the h5diff percentage option -p. PVN 2007/03/05 - Fixed a bug in h5dump that caused array indices greater than 2 ^32-1 not to be printed correctly - PVN 2007/2/19 Documentation ------------- F90 API ------- C++ API ------- - Changed StrType::StrType(const size_t& size); to StrType::StrType(const int dummy, const size_t& size); because the first one clashed with StrType::StrType(const hid_t existing_id); - BMR 2007/05/03 - Wrappers of H5Rcreate had incorrect prototypes. Added these overloaded functions: IdComponent::reference(void* ref, const char* name, DataSpace& dataspace, H5R_type_t ref_type = H5R_DATASET_REGION) IdComponent::void reference(void* ref, const char* name) IdComponent::reference(void* ref, const H5std_string& name) and will remove these incorrect member functions: DataSet::Reference DataType::Reference Group::Reference H5File::Reference - Added wrappers for H5Rdereference as constructors DataSet(IdComponent& obj, void* ref); Group(IdComponent& obj, void* ref); DataType(IdComponent& obj, void* ref); Platforms Tested ================ AIX 5.2 (32/64 bit) xlc 6.0.0.8 xlC 6.0.0.9 xlf 8.1.1.7 mpcc_r 6.0.0.8 mpxlf_r 8.1.1.7 AIX 5.3 (32/64 bit) xlc 7.0.0.0 xlC 7.0. xlf 9.1.0.3 Cray X1 water 3.0.35 Cray Standard C Version 5.4.0.7.4 Cray Fortran 5.4.0.7.3 Cray C++ 5.4.0.7.4 FreeBSD 6.2 gcc 3.4.6 g++ 3.4.6 HP-UX B.11.23 HP aC++/ANSI C B3910B A.06.02 HP F90 v3.1 (32-bit only) HP aC++/ANSI C B3910B A.06.02 IRIX64 6.5 (tesla -64) MIPSpro cc 7.4.2m F90 MIPSpro 7.4.2m C++ MIPSpro cc 7.4.2m IRIX64 6.5 (ucar mop1 64 & n32) MIPSpro cc 7.4.4m F90 MIPSpro 7.4.4m (64 only) C++ MIPSpro cc 7.4.4m Linux 2.4.20-28.7 gcc 2.96 (eirene, verbena) gcc 3.3.2 PGI compilers (pgcc, pgf90, pgCC) version 5.2-1 Absoft Fortran compiler v9.0 Intel(R) C++ 32-bit Version 8.1 Intel(R) Fortran 32-bit Version 8.1 MPIch 1.2.6 Linux 2.4.21-268-smp x86_64 gcc 3.3.1 (SuSE Linux, AMD) (mir) PGI 5.2-1 C and F90 (with k3-32) Intel(R) C++ 32-bit Version 8.1 Intel(R) Fortran 32-bit Version 8.1 Linux 2.6.16.46-0.12-debug #1 SMP ia64 GNU/Linux Intel(R) C++ Version 10.0.025 (ucar hir1) Intel(R) Fortran Itanium(R) Version 10.0.025 Linux 2.6.5-7.283-rtgfx Altix SMP ia64 Intel(R) C++ Version 9.0 (cobalt) Intel(R) Fortran Itanium(R) Version 9.0 SGI MPI OSF1 V5.1 (QSC) Compaq C V6.5-011 HP Fortran V5.5A-3548 Compaq C++ V6.5-036 MPIX200_64_r13.4 SunOS 5.8 32,46 Sun WorkShop 6 update 2 C 5.3 (Solaris 2.8) Sun WorkShop 6 update 2 Fortran 95 6.2 Sun WorkShop 6 update 2 C++ 5.3 SunOS 5.10 Sun WorkShop 6 update 2 C 5.3 Sun WorkShop 6 update 2 Fortran 95 6.2 Sun WorkShop 6 update 2 C++ 5.3 Patch 111685-13 Xeon Linux 2.4.21-32.0.1.ELsmp-perfctr-lustre (tungsten) Intel(R) C++ Version 9.0 Intel(R) Fortran Compiler Version 9.0 Xeon Linux 2.6.9-42.0.10.EL_lustre-1.4.10.1smp (abe) Intel(R) C++ Version 10.0 Intel(R) Fortran Compiler Version 10.0 Open MPI 1.2.2 IA-64 Linux 2.4.21.SuSE_292.til1 ia64 (NCSA tg-login) Intel(R) C++ Version 8.0 Intel(R) Fortran Compiler Version 8.0 mpich-gm-1.2.5..10-intel-r2 Windows XP Visual Studio 6.0 Visual Studio .NET (with Intel Fortran 9.0) Visual Studio 2005 (with Intel Fortran 9.0) cygwin (gcc 3.4.4) MAC OS X Intel Darwin 8.10.1 i686-apple-darwin8-gcc-4.0.1 (GCC) 4.0.1 Supported Configuration Features Summary ======================================== Key: y = tested and supported n = not supported or not tested in this release x = not working in this release dna = does not apply ( ) = footnote appears below second table Platform C F90 F90 C++ zlib SZIP parallel parallel IBM BG/L (15) n n n n y y Solaris2.8 64-bit y y y(1) y y y Solaris2.8 32-bit y y y(1) y y y Solaris2.9 64-bit y(1) y y(1) y y y Solaris2.9 32-bit y(1) y y(1) y y y Solaris2.10 64-bit y(1) y n y y y Solaris2.10 32-bit y(1) y n y y y IRIX64_6.5 64-bit y(2) y y y y y IRIX64_6.5 32-bit y(2) n n n y y HPUX11.00 y(1) y y y y y HPUX11.23-32bit n y n y y y HPUX11.23-64bit n n n y y y OSF1 v5.1 y y y y y y X1 y y y y y n AIX-5.1, 5.2 & 5.3 32-bit y y y y y y AIX-5.1, 5.2 & 5.3 64-bit y y y y y y Windows XP Visual Studio 6.0 n n n y y y Windows XP Visual Studio .NET n y(8) n y y y Windows XP Visual Studio 2005 n y(8) n y y y Windows XP Cygwin n n n y y y Mac OS X 10.3 n y(9) n y y y FreeBSD 6.2 n n n y y y RedHat 7.3 W (3) y(1) y(10) n y y y RedHat 7.3 W Intel (3) n y n y y y RedHat 7.3 W PGI (3) n y n y y y RedStorm (15) n y n y y n SuSe x86_64 gcc (3,13) n y(11) n y y y SuSe x86_64 icc (3,13) n y(13) n y y y Linux 2.4 Xeon C Lustre Intel (3,6) n y n y y y Linux 2.4 SuSE ia64 C Intel (3,7) y y y y y y Linux 2.4 SGI Altix ia64 Intel (3) y y y y y y Platform Shared static- Thread- STREAM- libraries(4) exec safe VFD IBM BG/L n y n n Solaris2.8 64-bit y x y y Solaris2.8 32-bit y x y y Solaris2.9 64-bit y x y y Solaris2.9 32-bit y x y y Solaris2.10 64-bit y x y y Solaris2.10 32-bit y x y y IRIX64_6.5 64-bit y y y y IRIX64_6.5 32-bit y y y y HPUX11.00 y x n y HPUX11.23 y y n y OSF1 v5.1 y y n y X1 n y n y AIX-5.1, 5.2 & 5.3 32-bit n y n y AIX-5.1, 5.2 & 5.3 64-bit n y n y Windows XP Visual Studio 6.0 y y n n Windows XP Visual Studio .Net y y n n Windows XP Visual Studio 2005 y y n n Windows XP Cygwin y y n n Mac OS X 10.3 y y n y FreeBSD 6.2 y y y y RedHat 7.3 W (3) y y y y RedHat 7.3 W Intel (3) n y n y RedHat 7.3 W PGI (3) n y n y RedStorm n y n y SuSe x86_64 gcc (3,13) n y n y SuSe x86_64 icc (3,13) y y(14) n y Linux 2.4 Xeon C Lustre Intel (3,6) y y n y Linux 2.4 SuSE ia64 C Intel (3,7) y y n n Linux 2.4 SGI Altix ia64 Intel (3) y y n y Compiler versions for each platform are listed in the "Platforms Tested" table found elsewhere in this file (RELEASE.txt). Unless otherwise noted, compilers used are the system compilers. Footnotes: (1) Using mpich 1.2.6 (2) Using mpt and mpich 1.2.6. (3) Linux 2.4 with GNU, Intel, and PGI compilers, as indicated. W or C indicates workstation or cluster, respectively. (4) Shared libraries are provided only for the C library, except on Windows where they are provided for C and C++. (5) Using mpt. (6) Linux 2.4.21-32.0.1. Xeon cluster with ELsmp-perfctr-lustre and Intel compilers (7) Linux 2.4.21, SuSE_292.til1. Ia64 cluster with Intel compilers (8) Requires Intel Fortran Compiler 9.0 integration (9) IBM XLF and Absoft (10) PGI, Absoft. No shared libraries with Absoft; use '--disable-shared'. (11) PGI and Intel compilers for both C and Fortran (12) AMD Opteron x86_64 (13) ifort (14) Yes with C and Fortran, but not with C++ (15) Only serial is ported. PFS does not work for PHDF5 yet. FMB/EIP - 2005/11/10 Known Problems ============== * QSC (an HP alpha-based OSF1 cluster) failed the testpar/testphdf5 sub-test "calloc". All other tests passed. This indicates dataset using chunked storage created by serial HDF5 may not work properly with parallel HDF5. The calloc test can be skipped by running "prun ... testphdf5 -x calloc". AKC - 2007/7/12. * Newer versions of g95 are not compatible with v1.6.5. Last known g95 version that is still compatiable with v1.6.5 is May 17, 2006. AKC - 2006/11/29. * Intel Compilers for Linux x86_86 platforms (EM64T-based, v8.1) has optimization error in the data types conversion code. Before running configure, edit the file config/intel-flags by changing the setting of PROD_CFLAGS from -O3 to -O0, then run configure. AKC - 2005/11/10. * When testing parallel HDF5 with the C compiler version MIPSpro 7.4.3 at IRIX 6.5, set enviroment variable MPI_TYPE_MAX to be a bigger number, for example 120000, in order to pass the complicated collective IO tests inside parallel HDF5 library. This is not a problem inside parallel HDF5 library. You can always set a bigger number in your system. KY - 2005/10/6 * A contiguous or chunked dataset created by a sequential version may not be modified with a parallel version of the library. Use the H5Pset_alloc_time function with H5D_ALLOC_TIME_EARLY to set up the dataset creation property list to avoid the problem. EIP - 2005/09/09 * The dataset created or rewritten with the v1.6.3 library or after can't be read with the v1.6.2 library or before when Fletcher32 EDC(filter) is enabled. There was a bug in the calculating code of the Fletcher32 checksum in the library before v1.6.3. The checksum value wasn't consistent between big-endian and little-endian systems. This bug was fixed in Release 1.6.3. However, after fixing the bug, the checksum value is no longer the same as before on little-endian system. The library release after 1.6.4 can still read the dataset created or rewritten with the library of v1.6.2 or before. SLU - 2005/7/8 * For version 6 (6.02 and 6.04) of the Portland Group compiler on AMD Opteron processor, there's a bug in the compiler for optimization(-O2). The library failed in several tests but all related to multi driver. The problem has been reported to the vendor. * test/big fails sometimes with the message "Possible overlap with another region." The test selects regions randomly, and this error occurs when two regions overlap each other; it is a bug in the test and not in HDF5. Since the error is triggered by a random situation, it will usually disappear if the test is re-run. * Newer SGI MIPSpro compilers (version 7.4.x) support C99 features but it has a "guard" statement in stdint.h that will #error and skip the rest of the header file if C99 option is not used explicitly. Hardset $CC to c99 will resolve the problem. AKC - 2004/12/13 * On IBM AIX systems, parallel HDF5 mode will fail some tests with error messages like "INFO: 0031-XXX ...". This is from the command poe. Set the environment variable MP_INFOLEVEL to 0 to minimize the messages and run the tests again. The tests may fail with messages like "The socket name is already in use". HDF5 does not use sockets (except for stream-VFD). This is due to problems of the poe command trying to set up the debug socket. Check if there are many old /tmp/s.pedb.* staying around. These are sockets used by the poe command and left behind due to failed commands. Ask your system administrator to clean them out. Lastly, request IBM to provide a means to run poe without the debug socket. * The h5dump tests may fail to match the expected output on some platforms (e.g. parallel jobs, Windows) where the error messages directed to "stderr" do not appear in the "right order" with output from stdout. This is not an error. * The stream-vfd test uses ip port 10007 for testing. If another application is already using that port address, the test will hang indefinitely and has to be terminated by the kill command. To try the test again, change the port address in test/stream_test.c to one not being used in the host. * The --enable-static-exec configure flag fails to compile for Solaris platforms. This is due to the fact that not all of the system libraries on Solaris are available in a static format. The --enable-static-exec configure flag also fails to correctly compile on IBM SP2 platform for the serial mode. The parallel mode works fine with this option. The --enable-static-exec configure flag also fails to correctly compile on the HPUX 11.00. It is suggested that you don't use this option on these platforms during configuration. * The Stream VFD is not supported under Windows. * Before building HDF5 F90 Library from source on Crays replace H5Aff.f90, H5Dff.f90 and H5Pff.f90 files in the fortran/src subdirectory in the top level directory with the Cray-specific files from the site: ftp://ftp.ncsa.uiuc.edu/HDF/HDF5/current/src/patches/ * Use --disable-shared configure flag if building with Absoft Fortran compiler. * Information about building with PGI and Intel compilers is available in INSTALL file sections 5.7 and 5.8. * In LANL QSC, the new cc compiler has problems converting small values of long long (absolute values less than 1**-308) to double. This triggers the test/dtypes to report failure in the Testing random sw long double -> double conversions If -ieee is used, the converted doubles spread over the range 0.0 to 1**-308. If -ieee is not used, the converted double values are mostly 0.0 but occasionally as 1**-308. This has been reported to the system staff. All other tests have passed. * On at least one system, SDSC DataStar, the scheduler (in this case LoadLeveler) sends job status updates to standard error when you run any executable that was compiled with the parallel compilers. This causes problems when running "make check" on parallel builds, as many of the tool tests function by saving the output from test runs, and comparing it to an exemplar. The best solution is to reconfigure the target system so it no longer inserts the extra text. However, this may not be practical. In such cases, one solution is to "setenv HDF5_Make_Ignore yes" prior to the configure and build. This will cause "make check" to continue after detecting errors in the tool tests. However, in the case of SDSC DataStar, it also leaves you with some 150 "failed" tests to examine by hand. A second solution is to write a script to run serial tests and filter out the text added by the scheduler. A sample script used on SDSC DataStar is given below, but you will probably have to customize it for your installation. Observe that the basic idea is to insert the script as the first item on the command line which executes the test. The script then executes the test and filters out the offending text before passing it on. #!/bin/csh set STDOUT_FILE=~/bin/serial_filter.stdout set STDERR_FILE=~/bin/serial_filter.stderr rm -f $STDOUT_FILE $STDERR_FILE ($* > $STDOUT_FILE) >& $STDERR_FILE set RETURN_VALUE=$status cat $STDOUT_FILE tail +3 $STDERR_FILE exit $RETURN_VALUE You get the HDF make files and test scripts to execute your filter script by setting the environment variable "RUNSERIAL" to the full path of the script prior to running configure for parallel builds. Remember to "unsetenv RUNSERIAL" before running configure for a serial build. Note that the RUNSERIAL environment variable exists so that we can prefix serial runs as necessary on the target system. On DataStar, no prefix is necessary. However, on an MPICH system, the prefix might have to be set to something like "/usr/local/mpi/bin/mpirun -np 1" to get the serial tests to run at all. In such cases, you will have to include the regular prefix in your filter script.