HDF5 version 1.9.18 currently under development ================================================================================ INTRODUCTION This document describes the differences between HDF5-1.9.0 and HDF5 1.9.x snapshot, and contains information on the platforms tested and known problems in HDF5-1.9.x. For more details check the HISTORY*.txt files in the HDF5 source. Links to HDF5 1.9.x source code can be found on THG's development FTP server at the following location: ftp://ftp.hdfgroup.uiuc.edu/pub/outgoing/hdf5/snapshots User documentation for the snapshot can be accessed directly at this location: http://www.hdfgroup.uiuc.edu/HDF5/doc_dev_snapshot/H5_dev/ For more information, see the HDF5 home page: http://www.hdfgroup.org/HDF5/ If you have any questions or comments, please send them to the HDF Help Desk: help@hdfgroup.org CONTENTS - New Features - Support for new platforms and languages - Bug Fixes since HDF5-1.8.0 - Platforms Tested - Known Problems New Features ============ Configuration: ------------- - Configuration suite now uses Autoconf 2.61, Automake 1.10.1, and Libtool 2.2.2 (MAM - 2008/05/05) - The new configure option "--disable-sharedlib-rpath" disables embedding the '-Wl,-rpath' information into executables when shared libraries are produced, and instead solely relies on the information in LD_LIBRARY_PATH. (MAM - 2008/05/15) Library: -------- - Added 'mounted' field to H5G_info_t struct. (QAK - 2008/07/15) Parallel Library: ----------------- Fortran Library: ---------------- C++ Library: ------------ - These member functions were added as wrapper for H5Rdereference to replace the incorrect IdComponent::dereference(). void H5Object::dereference(H5File& h5file, void* ref) void H5Object::dereference(H5Object& obj, void* ref) In addition, these constructors were added to create the associated objects by way of dereference: Attribute(H5Object& obj, void* ref); Attribute(H5File& file, void* ref); DataSet(H5Object& obj, void* ref); DataSet(H5File& file, void* ref); DataType(H5Object& obj, void* ref); DataType(H5File& file, void* ref); Group(H5Object& obj, void* ref); Group(H5File& obj, void* ref); (BMR - 2008/08/10) Tools: ------ - h5dump: added a printing of the compression ratio of uncompressed and compressed sizes for cases where compression filters are present. (PVN - 2008/05/01) - h5dump: added an option to allow a user defined formatting string for printf regarding floating point numbers. (PVN - 2008/05/06) - h5dump: support for external links, display the object that the external link points to. (PVN - 2008/05/12) - h5repack: add a userblock to an HDF5 file during the repack. (PVN - 2008/08/26) - h5repack: add 2 options that call H5Pset_alignment in the repacked file. (PVN - 2008/08/29) High-Level APIs: ------ Documentation ------------- Support for new platforms, languages and compilers. ======================================= Bug Fixes since HDF5-1.8.0 release ================================== Library ------- - Fixed pointer alignment violations that could occur during vlen conversion. (NAF - 2008/09/16) - Fixed problem where library could cause a segmentation fault when an invalid location ID was given to H5Giterate(). (QAK - 2008/08/19) - Fixed improper shutdown when objects have reference count > 1. The library now tracks reference count due to the application separately from that due to internal library routines. (NAF - 2008/08/19) - Fixed assertion failure caused by incorrect array datatype version. (NAF - 2008/08/08) - Fixed an issue where mount point traversal would fail when using multiple handles for the child. (NAF - 2008/08/07) - Fixed an issue where mount points were inaccessible when using multiple file handles for the parent. The mount table is now in the shared file structure (the parent pointer is still in the top structure). (NAF - 2008/08/07) - when an attribute was opened twice and data was written with one of the handles, the file didn't have the data. It happened because each handle had its own object structure, and the empty one overwrote the data with fill value. This is fixed by making some attribute information like the data be shared in the attribute structure. SLU - 2008/07/22 - Fixed issue where a group could have a file mounted on it twice. (QAK - 2008/07/15) - Fixed the problem with the searching of target file for H5Lcreate_external(). The searching pattern will depend on whether the target file's pathname is an absolute or a relative path. Please see the description in the RM for H5Lcreate_external(). (VC - 2008/04/08) - Fixed possible file corruption bug when encoding datatype descriptions for compound datatypes whose size was between 256 & 511 bytes and the file was opened with the "use the latest format" property enabled (with H5Pset_libver_bounds). (QAK - 2008/03/13) - Fixed bug in H5Aget_num_attrs() routine to handle invalid location ID correctly. (QAK - 2008/03/11) Configuration ------------- Performance ------------- Tools ----- - Fixed bug in h5ls that prevented relative group listings (like "h5ls foo.h5/bar") from working correctly (QAK - 2008/06/03) - Fixed bug in h5diff that prevented datasets & attributes with variable-length string elements from comparing correctly. (QAK - 2008/02/28) - h5import bug on Windows w/binary datasets. fread in windows needs a binary file to be open with 'rb' instead of 'r' otherwise it terminates execution if an end of file character is found on the input file. Besides that the binary file generated needs to be open with 'wb' , otherwise an end of line character is read twice. (PVN - 2008/02/19) - Fixed bug in h5dump that caused binary output to be made only for the first dataset, when several datasets were requested. (PVN - 2008/04/07) - h5dump: when doing binary output (-b), the stdout printing of attributes was done incorrectly. Removed printing of attributes when doing binary output. PVN - 2008/06/05 Documentation ------------- F90 APIs -------- C++ APIs -------- - Fixed bug that caused segfaults in Attribute::read. (BMR - 2008/04/20) - Fixed bug in PropList::getClassName to use portable HDfree instead of free. (BMR - 2008/04/20) - Fixed a design bug which allowed an Attribute object to create/modify attributes (bugzilla #1068). The API class hierarchy was revised to address the problem. Classes AbstractDS and Attribute are moved out of H5Object. Class Attribute now multiply inherits from IdComponent and AbstractDs and class DataSet from H5Object and AbstractDs. In addition, the data member IdComponent::id was moved into subclasses: Attribute, DataSet, DataSpace, DataType, H5File, Group, and PropList. (BMR - 2008/08/10) - IdComponent::dereference was incorrect and replaced as described in "New Features" section. (BMR - 2008/08/10) Platforms Tested ================ AIX 5.2 (32/64 bit) xlc 8.0.0.11 xlC 8.0 xlf 10.01.0000.0 mpcc_r 6.0.0.8 mpxlf_r 8.1.1.7 FreeBSD 6.2-STABLE i386 gcc 3.4.6 [FreeBSD] 20060305 (duty) g++ 3.4.6 [FreeBSD] 20060305 gcc 4.2.1 20080123 g++ 4.2.1 20080123 gfortran 4.2.1 20070620 FreeBSD 6.2-STABLE amd64 gcc 3.4.6 [FreeBSD] 20060305 (liberty) g++ 3.4.6 [FreeBSD] 20060305 gcc 4.2.1 20080123 g++ 4.2.1 20080123 gfortran 4.2.1 20080123 IRIX64 6.5 (64 & n32) MIPSpro cc 7.4.4m F90 MIPSpro 7.4.4m C++ MIPSpro cc 7.4.4m Linux 2.6.9 (RHEL4) Intel 10.0 compilers (abe.ncsa.uiuc.edu) Linux 2.4.21-47 gcc 3.2.3 20030502 (osage) Linux 2.6.9-42.0.10 gcc 3.4.6 20060404 (kagiso) PGI 7.0-7 (pgcc, pgf90, pgCC) Intel 9.1 (icc, ifort, icpc) Linux 2.6.16.27 x86_64 AMD gcc 4.1.0 (SuSE Linux), g++ 4.1.0, g95 (GCC 4.0.3) (smirom) PGI 6.2-5 (pgcc, pgf90, pgCC) Intel 9.1 (icc, iort, icpc) Linux 2.6.5-7.252.1-rtgfx #1 SMP ia64 Intel(R) C++ Version 9.0 (cobalt) Intel(R) Fortran Itanium(R) Version 9.0 SGI MPI 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 cc: Sun C 5.8 (linew) f90: Sun Fortran 95 8.2 CC: Sun C++ 5.8 Xeon Linux 2.4.21-32.0.1.ELsmp-perfctr-lustre (tungsten) gcc 3.2.2 20030222 Intel(R) C++ Version 9.0 Intel(R) Fortran Compiler Version 9.0 IA-64 Linux 2.4.21.SuSE_292.til1 ia64 (NCSA tg-login) gcc 3.2.2 Intel(R) C++ Version 8.1 Intel(R) Fortran Compiler Version 8.1 mpich-gm-1.2.5..10-intel-r2 Windows XP Visual Studio .NET Visual Studio 2005 w/ Intel Fortran 9.1 Cygwin(native gcc compiler and g95) MinGW(native gcc compiler and g95) Windows XP x64 Visual Studio 2005 w/ Intel Fortran 9.1 Windows Vista Visual Studio 2005 MAC OS 10.4 (Intel) gcc i686-apple-darwin8-gcc-4.0.1 (GCC) 4.0.1 G95 (GCC 4.0.3 (g95 0.91!) Nov 21 2006) Alpha Open VMS 7.3 Supported Configuration Features Summary ======================================== In the tables below y = tested and supported n = not supported or not tested in this release x = not working in this release ( ) = footnote appears below second table = testing incomplete on this feature or platform W or C indicates workstation or cluster, respectively. Platform C F90 F90 C++ zlib SZIP parallel parallel SunOS5.8 64-bit n y n y y y SunOS5.8 32-bit n y n y y y SunOS5.10 64-bit n y n y y y SunOS5.10 32-bit n y n y y y IRIX64_6.5 64-bit n y y y y y IRIX64_6.5 32-bit n n n n y y AIX-5.2 32-bit y y y y y y AIX-5.2 64-bit y y y y y y Cray XT3 (not tested for this release) n n n n n n Windows XP n y(3) n(3) y y y Windows XP x64 n y(3) n(3) y y y Windows Vista n n n y y y Mac OS X 10.4 PowerPC n n Mac OS X 10.4 Intel n y n y y y FreeBSD 6.2 32-bit n n n y y y FreeBSD 6.2 64-bit RedHat EL4 2.6.9 i686 GNU W y(2) y(4) y(2) y y y RedHat EL4 2.6.9 i686 Intel W n y n y y n RedHat EL4 2.6.9 i686 PGI W n y n y y n SuSe Linux 2.6.16 x86_64 GNU (5) W y(2) y n y y y SuSe Linux 2.6.16 x86_64 Int (5) W n y n y n n SuSe Linux 2.6.16 x86_64 PGI (5) W n y n y n n RHL9 Linux 2.4 Xeon Lustre Intel C n y n y y n RHEL3 Linux 2.4 Xeon Intel W n y n n y n RHEL4 Linux 2.6 Xeon Lustre Int C n y n y y n SuSE Linux 2.4 ia64 Intel C y(1) y y y y y SuSe Linux 2.6.5 SGI Altix ia64 Intel C n y n y n y Alpha OpenVMS 7.3.2 n y n y n n Platform Shared Shared Shared static- Thread- C libs F90 libs C++ libs exec safe SunOS 5.8 32-bit y y y x y SunOS 5.8 64-bit y y y x y SunOS 5.10 32-bit y y y x y SunOS 5.10 64-bit y y y x y IRIX64_6.5 32-bit y n y x y IRIX64_6.5 64-bit y y n x y AIX-5.2 32-bit n n n x n AIX-5.2 64-bit n n n x n Cray XT3 (not tested for this release) n n n x n Windows XP y y(3) y y y Windows XP x64 y y(3) y y y Windows Vista y n n y y Mac OS X 10.4 Intel Duo y y y x n FreeBSD 6.2 32-bit y y y x n FreeBSD 6.2 64-bit y y y x n RHEL4 2.6.9 i686 GNU W y y(4) y x y RHEL4 2.6.9 i686 Intel W y y y x n RHEL4 2.6.9 i686 PGI W y y y x n SuSE Linux 2.6.16 x86_64 GNU (5) W y y y x y SuSE Linux 2.6.16 x86_64 Intel(5) W y y y x n SuSE Linux 2.6.16 x86_64 PGI(5) W y y y x n RHL9 Linux 2.4 Xeon Lustre Intel C y y y x n RHEL3 Linux 2.4 Xeon Intel W y n n x n RHEL4 Linux 2.6 Xeon Lustre Intel C y y y x n SuSE Linux 2.4 ia64 Intel C y y y x n SuSe Linux 2.6.5 SGI Altix ia64 Intel C n n n x n Notes: (1) Using mpich 1.2.6. (2) Using mpich2 1.0.6. (3) Using Visual Studio 2005 or Cygwin (4) With PGI and Absoft compilers. (5) AMD Opteron x86_64 Compiler versions for each platform are listed in the preceding "Platforms Tested" table. Known Problems ============== * We discovered two problems when running collective IO parallel HDF5 tests with chunking storage on ChaMPIon MPI compiler on tungsten, a linux cluster at NCSA. Under some complex selection cases, 1) MPI_Get_element returns the wrong value. 2) MPI_Type_struct also generates wrong derived data type and corrupt data may be generated. This only happens when turning on collective IO with chunking storage with some complex selections. We haven't found these problems on other MPI-IO compilers. If you encounter these problems, you may use Independent IO instead. Change the following line in your code: H5Pset_dxpl_mpio(xfer_plist, H5FD_MPIO_COLLECTIVE); to H5Pset_dxpl_mpio(xfer_plist, H5FD_MPIO_INDEPENDENT); KY - 2007/08/24 * For SNL, spirit/liberty/thunderbird: The serial tests pass but parallel tests failed with MPI-IO file locking message. AKC - 2007/6/25. * On Intel 64 Linux cluster (RH 4, Linux 2.6.9) with Intel 10.0 compilers use -mp -O1 compilation flags to build the libraries. Higher level of optimization causes failures in several HDF5 library tests. * For HPUX 11.23 many tools tests failed for 64-bit version when linked to the shared libraries (tested for 1.8.0-beta2) * For SNL, Red Storm: only paralle HDF5 is supported. The serial tests pass and the parallel tests also pass with lots of non-fatal error messages. * For LLNL, uP: both serial and parallel pass. Zeus: serial passes but parallel fails with a known proglem in MPI. ubgl: serial passes but parallel fails. * on SUN 5.10 C++ test fails in the "Testing Shared Datatypes with Attributes" test * configuring with --enable-debug=all produces compiler errors on most platforms. Users who want to run HDF5 in debug mode should use --enable-debug rather than --enable-debug=all to enable debugging information on most modules. * On Mac OS 10.4, test/dt_arith.c has some errors in conversion from long double to (unsigned) long long and from (unsigned)long long to long double. * On Altix SGI with Intel 9.0 testmeta.c would not compile with -O3 optimization flag. * On VAX, Scaleoffset filter isn't supported. The filter cannot be applied to HDF5 data generated on VAX. Scaleoffset filter only supports IEEE standard for floating-point data. * On Cray X1, a lone colon on the command line of h5dump --xml (as in the testh5dumpxml.sh script) is misinterpereted by the operating system and causes an error. * On mpich 1.2.5 and 1.2.6, we found that if more than two processes contribute no IO and the application asks to do IO with collective, we found that when using 4 processors, a simple collective write will be hung sometimes. This can be verified with t_mpi test under testpar. * On IRIX6.5, when C compiler version >7.4, the complicate MPI derived data type code will work. However, the user should be aware to enlarge MPI_TYPE_MAX environment variable to some certian value in order to use collective irregular selection code. For example, the current parallel HDF5 test needs to enlarge MPI_TYPE_MAX to 200,000 to make the test pass. * 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 - 2005/6/30 * For the version 6(6.02 and 6.04) of 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. * 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 mean to run poe without the debug socket. * The C++ library's tests fails when compiling with PGI C++ compiler. The workaround until the problem is correctly handled is to use the flag "--instantiate=local" prior to the configure and build steps, as: setenv CXX "pgCC --instantiate=local" for pgCC 5.02 and higher * 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. It is suggested that you don't use this option on these platforms during configuration. * With the gcc 2.95.2 compiler, HDF 5 uses the `-ansi' flag during compilation. The ANSI version of the compiler complains about not being able to handle the `long long' datatype with the warning: warning: ANSI C does not support `long long' This warning is innocuous and can be safely ignored. * The Stream VFD was not tested yet under Windows. It is not supported in the TFLOPS machine. * The ./dsets tests failed in the TFLOPS machine if the test program, dsets.c, is compiled with the -O option. The hdf5 library still works correctly with the -O option. The test program works fine if it is compiled with -O1 or -O0. Only -O (same as -O2) causes the test program to fail. * Certain platforms give false negatives when testing h5ls: - Cray J90 and Cray T90IEEE give errors during testing when displaying some floating-point values. These are benign differences due to the different precision in the values displayed and h5ls appears to be dumping floating-point numbers correctly. * Not all platforms behave correctly with szip's shared libraries. Szip is disabled in these cases, and a message is relayed at configure time. Static libraries should be working on all systems that support szip, and should be used when shared libraries are unavailable. There is also a configure error on Altix machines that incorrectly reports when a version of szip without an encoder is being used. * On some platforms that use Intel and Absoft compilers to build HDF5 fortran library, compilation may fail for fortranlib_test.f90, fflush1.f90 and fflush2.f90 complaining about exit subroutine. Comment out the line IF (total_error .ne. 0) CALL exit (total_error) * Information about building with PGI and Intel compilers is available in INSTALL file sections 5.7 and 5.8 * 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 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 scipts 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 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. * H5Ocopy() does not copy reg_ref attributes correctly when shared-message is turn on. The value of the reference in the destination attriubte is wrong. This H5Ocopy problem will affect h5copy tool