HDF5 version 1.7.52 currently under development ================================================================================ INTRODUCTION This document describes the differences between HDF5-1.6.* and HDF5-1.7.*, and contains information on the platforms tested and known problems in HDF5-1.7.*. 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/ 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 and languages - Bug Fixes since HDF5-1.6.0 - Platforms Tested - Known Problems New Features ============ Configuration: -------------- - Added --enable-build-all option, which only developers should need to use. -JML 2005/10/24 - Configure uses the 'TR' variable to let the user override the path to the 'tr' utility. -JML 2005/10/17 - Configure can recognize -lmpich as a form of MPI library. -AKC- 2005/9/28. - MD5 checksumming has been added to snapshot releases. Release tarballs will be accompanied by .md5 checksum files, which can be verified using the md5sum utility. -JML 2005/9/6 - Some configure flags are incompatible (e.g., the C++ APIs cannot be built using the parallel version of HDF5). configure will now output errors when some common incompatible features are used together. -JML 2005/9/6 - A new API function, H5Tis_hard(), was added to the library. It checks if a conversion function is a compiler (hard) conversion. SLU - 2005/9/6 - t_mpi will run the test_mpio_derived_dtype by default unless it is known not working (indicated by macro H5_MPI_COMPLEX_DERIVED_DATATYPE_WORKS not defined.) -AKC- 2005/8/23. - Test execution has changed in a number of ways: When make is invoked in parallel (using -j), sequential tests are now executed as parallel make targets. This should make them finish more quickly on machines with multiple processors. Since test output is garbled when they are executed by parallel make, tests now dump their output to foo.log files and foo.logsh files (for test scripts). These logs are printed to the screen only when a test fails or when all tests in the current directory have completed successfully. When tests pass, they will create a foo.chkexe file. This prevents the test from executing again until the test or main library changes. All files generated by tests (*.chkexe, *.log, and any *.h5 files created) can be removed by invoking 'make check-clean'. Sequential and parallel library tests can now be invoked separately. 'make check-s' will execute only sequential tests, and 'make check-p' will execute only parallel tests. 'make check' will still execute all tests. -JML 2005/08/03 - On windows, all.zip is deprecated. users should read INSTALL_Windows.txt to know the details. Reasons to deprecate all.zip: 1. Avoid confliction for windows programmers 2. Decrease size of CVS tree by adding all.zip 3. Avoid using winzip as the intermediate step --KY 2005/04/22 - When HDF5 is created as a shared library, it now uses libtool's shared library versioning scheme. -JML 2005/04/18 - HDF5 now uses automake 1.9.5 to generate Makefiles.in. This has a number of effects on users: The Fortran compiler should be set using the environment variable $FC, not $F9X. F9X still works, but is depreciated. The output of make may be different. This should be only a cosmetic effect. make depened (or make dep) is no longer recognized, since automake handles dependency tracking. Some new configure options exist. --enable-dependency-tracking and --disable-dependency-tracking are used to control automake's dependency tracking. Dependencies are on by default *on most platforms and compilers*. If --enable-dependency-tracking is used, they will be enabled on any platform. However, this can slow down builds or even cause build errors in some cases. Likewise, --disable-dependency-tracking can speed up builds and avoid some build errors. Some make targets have alternate names. make check-install and make installcheck do the same thing, for instance. pmake on IRIX can be invoked from the root directory, but the -V flag must be used to invoke it in any subdirectory or it will give an error about undefined variables. JML 2005/01 - 2005/03 - Hardware conversion between long double and integers is also added. SLU 2005/02/10 - Started to support software conversion between long double and integers. Hardware conversion will come very soon. SLU - 2005/1/6 - Intel v8.0 compiler would infinite loop when compiling some test code with -O3 option. Changed enable-production default compiler option to -O2. AKC - 2004/12/06 - Long double is assumed to be a supported C data type. It is a stanadard C89 type. AKC - 2004/10/22 - The IA64 will use ecc as the C++ compiler by default. - Added some initial support for making purify (or similar memory checking products) happier by initializing buffers to zero and disabling the internal free list code. To take advantage of this, define 'H5_USING_PURIFY' in your CFLAGS when building the library. QAK - 2004/07/23 - Fixed the long compile time of H5detect.c when v7.x Intel Compiler is used with optimization NOT off. AKC - 2004/05/20 - Fixed configure setting of C++ for OSF1 platform. AKC - 2004/01/06 - Prefix default is changed from /usr/local to `pwd`/hdf5. AKC - 2003/07/09 Library: -------- - Added HSYS_ERROR which retrieves the system error message and pushes it to the error stack. This gives more information of the failed system call. AKC - 2005/08/04 - Added H5F_OBJ_LOCAL flag to H5Fget_obj_count() & H5Fget_obj_ids(), to allow querying for objects in file that were opened with a particular file ID, instead of all objects opened in file with any file ID. QAK - 2005/06/01 - Added H5T_CSET_UTF8 character set to mark datatypes that use the UTF-8 Unicode character encoding. Added tests to ensure that library handles UTF-8 object names, attributes, etc. -JL 2005/05/13 - HDF5 supports collective MPI-IO for irregular selection with HDF5 dataset. Irregular selection is when users use H5Sselect_hyperslab more than once for the same dataset. Currently, not all MPI-IO packages support complicated MPI derived datatype used in the implementation of irregular selection INSIDE HDF5. 1) DEC 5.x is not supporting complicated derived datatype. 2) For AIX 5.1, if your poe version number is 3.2.0.19 or lower, please edit powerpc-ibm-aix5.x under hdf5/config, Find the line with << hdf5_mpi_complex_derived_datatype_works>> and UNCOMMENT this line before the configure. check poe version with the following command: lpp -l all | grep ppe.poe 3) For Linux cluster, if mpich version is 1.2.5 or lower, collective irregular selection IO is not supported, internally independent IO is used. 4) For IRIX 6.5, if C compiler version is 7.3 or lower, collective irregular selection IO is not supported, internally independent IO is used. 5) For platforms which internally used mpich, if the mpich version is 1.2.5 or lower, please find the corresponding config file and add hdf5_mpi_complex_derived_datatype_works='no' at the end of the configuration file. For example, at NCSA SGI Altix, the internal mpich library is 1.2.5. So hdf5_mpi_complex_derived_datatype_works='no' should be added at the end of the config file ia64-linux-gnu. KY - 2005/09/12 - HDF5 N-bit filter HDF5 support N-bit filter from this version, The N-Bit filter is used effectively for compressing data of N-Bit datatype as well as compound and array datatype with N-Bit fields. KY - 2005/04/15 - HDF5 scaleoffset filter HDF5 supports scaleoffset filter for users to do data compression through HDF5 library. Scale-Offset compression performs a scale and/or offset operation on each data value and truncates the resulting value to a minimum number of bits and then stores the data. Scaleoffset filter supports floating-point and integer datatype. Please check the HDF5 reference manual for this. KY - 2005/06/06 - Retired SRB vfd (--with-srb). Functions H5Pset_fapl_srb and H5Pget_fapl_srb were removed. EIP - 2005/04/07 - Retired GASS vfd (--with-gass). Functions H5Pset_fapl_gass and H5Pget_fapl_gass are removed too. AKC - 2005/3/3 - Pablo was removed from the source code EIP - 2005/01/21 - Modified registration of SZIP to dynamically detect the presence or absence of the encoder. Changed configure and Makefiles, and tests to dynamically detect encoder. BEM - 2004/11/02 - Added function H5Pget_data_transform, together with the previously added H5Pset_data_transform, to support the data transform feature. AKC - 2004/10/26 - Compound datatype has been enhanced with a new feature of size adjustment. The size can be increased and decreased(without cutting the last member) as long as it doesn't go down to zero. No API change is involved. SLU - 2004/10/1 - Put back 6 old error API functions to be backward compatible with version 1.6. They are H5Epush, H5Eprint, H5Ewalk, H5Eclear, H5Eset_auto, H5Eget_auto. Their new equivalent functions are called H5Epush_stack, H5Eprint_stack, H5Ewalk_stack, H5Eclear_stack, H5Eset_auto_stack, H5Eget_auto_stack. SLU - 2004/9/2 - 4 new API functions, H5Tencode, H5Tdecode, H5Sencode, H5Sdecode were added to the library. Given object ID, these functions encode and decode HDF5 objects(data type and space) information into and from binary buffer. SLU - 2004/07/21 - Modified the way how HDF5 calculates 'pixels_per_scanline' parameter for SZIP compression. Now there is no restriction on the size and shape of the chunk except that the total number of elements in the chunk cannot be bigger than 'pixels_per_block' parameter provided by the user. EIP - 2004/07/21 - Added support for SZIP without encoder. Added H5Zget_filter_info and changed H5Pget_filter and H5Pget_filter_by_id to support this change. JL/NF - 2004/06/30 - SZIP always uses K13 compression. This flag no longer needs to be set when calling H5Pset_szip. If the flag for CHIP compression is set, it will be ignored (since the two are mutually exclusive). JL/NF - 2004/6/30 - A new API function H5Fget_name was added. It returns the name of the file by object(file, group, data set, named data type, attribute) ID. SLU - 2004/06/29 - Added support for user defined identifier types. NF/JL - 2004/06/29 - A new API function H5Fget_filesize was added. It returns the actual file size of the opened file. SLU - 2004/06/24 - New Feature of Data transformation is added. AKC - 2004/05/03. - New exception handler for datatype conversion is put in to replace the old overflow callback function. This exception handler is set through H5Pset_type_conv_cb function. SLU - 2004/4/27 - Added option that if $HDF5_DISABLE_VERSION_CHECK is set to 2, will suppress all library version mismatch warning messages. AKC - 2004/4/14 - A new type of dataspace, null dataspace(dataspace without any element) was added. SLU - 2004/3/24 - Data type conversion(software) from integer to float was added. SLU - 2004/3/13 - Data type conversion(software) from float to integer was added. Conversion from integer to float will be added later. SLU -2004/2/4 - Added new H5Premove_filter routine to remove I/O pipeline filters from dataset creation property lists. PVN - 2004/01/26 - Added new 'compare' callback parameter to H5Pregister & H5Pinsert routines. QAK - 2004/01/07 - Data type conversion(hardware) between integers and floats was added. SLU 2003/11/21 - New function H5Iget_file_id() was added. It returns file ID given an object(dataset, group, or attribute) ID. SLU 2003/10/29 - Added new fields to the H5G_stat_t for more information about an object's object header. QAK 2003/10/06 - Added new H5Fget_freespace() routine to query the free space in a given file. QAK 2003/10/06 - Added backward compatability with v1.6 for new Error API. SLU - 2003/09/24 - Changed 'objno' field in H5G_stat_t structure from 'unsigned long[2]' to 'haddr_t'. QAK - 2003/08/08 - Changed 'fileno' field in H5G_stat_t structure from 'unsigned long[2]' to 'unsigned long'. QAK - 2003/08/08 - Changed 'hobj_ref_t' type from structure with array field to 'haddr_t'. QAK - 2003/08/08 - Object references (hobj_ref_t) can now be compared with the 'objno' field in the H5G_stat_t struct for testing if two objects are the same within a file. QAK - 2003/08/08 - Switched over to new error API. SLU - 2003/07/25 Parallel Library: ----------------- - Added mpich2 as a testing "platform" informally. AKC - 2005/9/28. - A dataset created in serial mode with H5D_ALLOC_TIME_INCR allocation setting was not extendible, either explicitly by H5Dextend or implicitly by writing to unallocated chunks. Library now allocates more space when needed or directed if the file is opened by parallel mode, independent of what the dataset allocation mode is. CC/AKC - 2005/08/29. - Allow compressed, chunked datasets to be read in parallel. QAK - 2004/10/04 - Add options of using atomicity and file-sync to test_mpio_1wMr. AKC - 2003/11/13 - Added parallel test, test_mpio_1wMr, which tests if the underlaying parallel I/O system is conforming to the POSIX write/read requirement. AKC - 2003/11/12 Fortran Library: ---------------- - added support for shared Fortran libraries. -JML 2005/09/20 - added missing h5tget_member_class_f function EIP 2005/04/06 - added new functions h5fget_name_f and h5fget_filesize_f EIP 2004/07/08 - h5dwrite/read_f and h5awrite/read_f functions only accept dims parameter of the type INTEGER(HSIZE_T). - added support for native integers of 8 bytes (i.e. when special compiler flag is specified to set native fortran integers to 8 bytes, for example, -i8 flag for PGI and Absoft Fortran compilers, -qintsize=8 flag for IBM xlf compiler). EIP 2005/06/20 - added support for "big" REAL and DOUBLE PRECISION types (usually the size is specified by compilers flags like -r8, -r16, etc.) Known problem: multi file test fails when REAL is 16 bytes. EIP 2005/09/8 C++ Library: ---------------- - added support for shared C++ libraries. -JML 2005/09/20 - Added missing member functions H5::CompType::getMemberArrayType H5::CompType::getMemberVarLenType H5::AbstractDs::getArrayType H5::AbstractDs::getVarLenType H5::CommonFG::openArrayType H5::CommonFG::openVarLenType H5::PropList::copyProp -- this will replace the current H5::PropList::copyProp in later releases due to incorrect prototype. H5::IdComponent::getHDFObjType BMR - 2005/08/08 Tools: ------ - Removed obsolete pdb2hdf5 tool from tools/misc -JML 2005/10/24 - Added build_h5perf_alone.sh that builds h5perf by standalone mode. AKC - 2005/09/18. - Sped up h5dump on files with large numbers of objects. QAK - 2005/08/25 - Added a standalone mode for building h5perf. AKC - 2005/08/12 - new tool, h5jam. See reference manual. 2004/10/08 - h5repack.sh did not report errors encountered during tests. It does now. AKC - 2004/04/02 - Added the MPI-I/O and MPI-POSIX drivers to the list of VFL drivers available for h5dump and h5ls. RPM & QAK - 2004/02/01 - Added option --vfd= to h5ls to allow a VFL driver to be selected by a user. RPM & QAK - 2004/02/01 - Added option -showconfig to compiler tools (h5cc,h5fc,h5c++). AKC - 2004/01/08 - Install the "h5cc" and "h5fc" tools as "h5pcc" and "h5pfc" respectively if library is built in parallel mode. WCW - 2003/11/04 - Added metadata benchmark (perform/perf_meta). SLU - 2003/10/03 - Changed output of "OID"s from h5dump from "-" to ":::" to ":" QAK - 2003/08/08 High-Level APIs: ------ - Added Packet Table API for creating tables with less overhead than H5TB API. Added C++ wrapper for Packet Tables. See documentation. JML - 2004/03/28 Support for new platforms, languages and compilers. ======================================= - Added support for Cray X1. JML - 2005/10/03 - PGI Fortran compiler is supported on Linux64 systems (x86_64) EIP - 2004/08/19 - Absoft compiler f95 v9.0 supported on Linux 2.4 EIP - 2004/07/29 - HDF5 Fortran APIs are supported on Mac OSX with IBM XL Fortran compiler version 8.1. This is a default compiler. - HDF5 Fortran APIs are supported on MAC OSX with Absoft F95 compiler version 8.2; set F9X environment varibale to f95, for example setenv F9X f95 Use --disable-shared --enable-static configure flags when Absoft compiler is used. EIP - 2004/07/27 - HDF5 Fortran APIs are supported on MAC OSX with IBM XL Fortran Compiler version 8.1 Use "--disable-shared --enable-static" configure flags along with the "--enable-fortran" flag to build Fortran library. EIP - 2004/01/07 Bug Fixes since HDF5-1.6.0 release ================================== Library ------- - Fixed core dump when closing root groups opened through two different file handles that operate on the same actual file. QAK - 2005/10/02 - Corrected errors when performing various operations on a group opened by dereferencing an object reference. QAK - 2005/07/30 - Fixed a bug with named datatypes where a copy of a named datatype used to create a dataset would accidentally use the original named datatype for the dataset's datatype. QAK - 2005/07/23 - Made H5Fget_name() be consistent and always return name of actual file the ID is in. (Instead of the name of the top file in a file mounting hierarchy). QAK - 2005/07/19 - Reworked internal file mounting semantics to hopefully eliminate mounting problems. We now require that files that are mounting together all have the same "file close degree". QAK - 2005/07/19 - More bug fixes on holding open files that are mounted and have IDs open. QAK - 2005/07/14 - Don't unmount child files until the parent file actually closes. (Previously, if an object is holding open a file, the child files would get unmounted too early). QAK - 2005/07/05 - Fixed bug where unmounted files could cause the library to go into an infinite loop when shutting down. QAK - 2005/06/30 - The library didn't save the information of family driver in file. The original file member size was lost after file was closed (see bug #213). This has been fixed by saving driver name and member file size in the superblock. SLU - 2005/6/24 - Fixed bug with hyperslab selections that use selection offsets and operate on chunked datasets going into infinite loop or dumping core. QAK - 2005/06/17 - Corrected memory leak and possible corruption when opening a group. QAK - 2005/06/17 - Added check for opaque datatype tags being too long (check against H5T_OPAQUE_TAG_MAX, currently set to 256). QAK - 2005/06/14 - Fixed various errors in maintaining names for open objects in the face of unusual mount & unmount operations. QAK - 2005/06/08 - "SEMI" and "STRONG" file close degree settings now apply only to the particular file ID being closed, instead of operating on all open file IDs for a given file. QAK - 2005/06/01 - For family driver, the library didn't save member size in file. When file is reopened, the size of 1st member file determine the member size. Now member size is saved in file and is used to define member file size. Wrong file access property of member size will result in a failure. Using any other driver except family will cause library to return error. So is multi driver. SLU - 2005/05/24 - Fixed error in opening object in group that was opened in mounted file which has been unmounted. QAK - 2005/03/17 - Fixed a racing condition in MPIPOSIX virtual file drive close function. Now all processes must completed the close before any of them is returned. This prevents some "faster" processes start accessing the file for another purpose (e.g., open with truncate) while other "slower" processes have not closed the same file with the previous purpose. AKC - 2005/03/01 - H5Tget_member_value calls for enum datatype didn't return correct value if H5Tenum_valueof was called first. It's fixed. SLU - 2005/02/08 - For variable-length string, H5Tget_class returned H5T_STRING as its class. But H5Tdetect_class and H5Tget_member_class considered it as H5T_VLEN. This is fixed to let all these 3 functions treat it as H5T_STRING. SLU - 2005/02/08 - The byte order of 1-byte integer types was fixed as little endian even on a big-endian machine. This has been corrected. SLU - 2005/02/07 - Fix segmentation fault when calling H5Fflush with an attribute that hasn't had a value written to it open. QAK - 2004/10/18 - Back up supporting bitfield and time types in H5Tget_native_type. Leave it to future support. The function simply returns error message of "not support" for bitfield and time types. SLU - 2004/10/5 - Fixed address check in Core VFL driver to avoid spurious address/size overflows for odd valued addresses and/or sizes. QAK - 2004/09/27 - Fixed parallel bug in which some processes attempted collective I/O while others did independent I/O. Bug appeared when some processes used point selections, and others didn't. JRM - 2004/9/15 - Corrected error where dataset region references were written in an incorrect way on Cray machines. PVN & QAK - 2004/09/13 - The H5Tget_native_type now determines the native type for integers based on the precision. This is to avoid cases of wrongly converting an int to a short in machines that have a short of 8 bytes but with 32bit precision (e.g Cray SV1). PVN - 2004/09/07 - Changed H5Dread() to not overwrite data in an application's buffer with garbage when accessing a chunked dataset with an undefined fill value and an unwritten chunk is uncountered. QAK - 2004/08/25 - Fixed error which could cause a core dump when a type conversion routine was registered after a compound datatype had been converted and then an equivalment compound datatype was converted again. QAK - 2004/08/07 - Fixed memory overwrite when encoding "multi" file driver information for file's superblock. QAK - 2004/08/05 - Fixed obscure bug where a filter which failed during chunk allocation could allow library to write uncompressed data to disk but think the data was compressed. QAK - 2004/07/29 - Fixed bug where I/O to an extendible chunked dataset with zero-sized dimensions would cause library to fail an assertion. QAK - 2004/07/27 - Fixed bug where chunked datasets which have filters defined, allocation time set to "late" and whose chunks don't align with the dataspace bounds could have incorrect data stored when overwriting the entire dataset on the first write. QAK - 2004/07/27 - Added check to ensure that dataspaces have extents set. JML-2004/07/26 - Fixed bug on some Solaris systems where HDF5 would try to use gettimeofday() when that function didn't work properly. JML - 2004/07/23 - Fixed bug in H5Sset_extent_simple where setting maximum size to non-zero, then to zero would cause an error. JML - 2004/07/20 - Allow NULL pointer for buffer parameter to H5Dread & H5Dwrite when not writing data ("none" selection or hyperslab or point selection with no elements defined). QAK - 2004/07/20 - Calling H5Gcreate() on "/" or "." throws an error instead of failing quietly. JML - 2004/07/19 - Fixed bug where setting file address size to be very small could trigger an assert if the file grew to more than 64 KB. Now throws an error and data can be recovered. JL/NF - 2004/07/14 - Fixed bug where "resurrecting" a dataset was failing. QAK - 2004/07/14 - Fixed bug where incorrect data could be read from a chunked dataset after it was extended. QAK - 2004/07/12 - Fixed failure to read data back from file of compound type with variable-length string as field. SLU - 2004/06/10 - Fixed potential file corruption bug when a block of metadata could overlap the end of the internal metadata accumulator buffer and the buffer would be extended correctly, but would incorrectly change it's starting address. QAK - 2004/06/09 - Opaque datatype with no tag failed for some operations. Fixed. SLU - 2004/6/3 - Fixed potential file corruption bug where dimensions that were too large (a value greater than could be represented in 32-bits) could cause the incorrect amount of space to be allocated in a file for the raw data for the dataset. QAK - 2004/06/01 - Fixed dtypes "sw long double -> double" failure in QSC class machines. AKC - 2004/4/16 - Fixed problem with fletcher32 filter when converting data of different endianess. PVN - 2004/03/10 - Fixed problem with H5Tget_native_type() not handling opaque fields correctly. QAK - 2004/01/31 - Fixed several errors in B-tree deletion code which could cause a B-tree (used with groups and chunked datasets) to become corrupt with the right sequence of deleted objects. QAK - 2004/01/19 - Fixed small internal memory leaks of fill-value information. QAK - 2004/01/13 - Fixed bug that caused variable-length datatypes (strings or sequences) used for datasets in files with objects that were unlinked to fail to be read/written to a file. QAK - 2004/01/13 - Detect situation where szip 'pixels per block' is larger than the fastest changing dimension of a dataset's chunk size and disallow this (due to limits in szip library). QAK - 2003/12/31 - Fixed bug with flattened hyperslab selections that would generate incorrect hyperslab information with certain high-dimensionality combinations of start/stride/count/block information. QAK - 2003/12/31 - Fixed bug with variable-length datatypes used in compound datatypes. SLU - 2003/12/29 - Fixed bug in parallel I/O routines that would cause reads from "short datasets" (datasets which were only partially written out) to return invalid data. QAK & AKC - 2003/12/19 - Fixed bug where scalar dataspaces for attributes were reporting as simple dataspaces. QAK - 2003/12/13 - Fixed problem with selection offsets of hyperslab selections in chunked datasets causing the library to go into an infinite loop. QAK - 2003/12/13 - Fixed H5Giterate to avoid re-using index parameter after iteration callback has been called (allows iteration callback to modify the index parameter itself). QAK - 2003/12/06 - Fixed various floating-point conversion problems, including a change which could corrupt data when converting from double->float. QAK - 2003/11/24 - Changed "single process" metadata writing in library to collective I/O by all processes, in order to guarantee correct data being written with MPI-I/O. QAK - 2003/11/20 - Fixed problems with fill values and variable-length types and also I/O on VL values that were set to NULL. QAK - 2003/11/08 - Fixed problems with MPI datatypes that caused ASCI Q machine to hang. QAK - 2003/10/28 - Removed HDF5_MPI_PREFER_DERIVED_TYPES environment variable support, since it had no benefit. QAK - 2003/10/28 - Single hyperslab selections (which were set with only one call to H5Sselect_hyperslab) that had dimensions that could be "flattened" but were interspersed with dimensions that could not be flattened were not correctly handled, causing core dumps. QAK - 2003/10/25 - Fixed incorrect datatype of the third parameter to the Fortran90 h5pset(get)_cache_f subroutine (INTEGER to INTEGER(SIZE_T)) EIP - 2003/10/13 - Fixed problems with accessing variable-length data datatypes on Crays. QAK - 2003/10/10 - Fixed potential file corruption bug when too many object header messages (probably attributes, from a user perspective) were inserted into an object header and certain other conditions were met. QAK - 2003/10/08 - Changed implementation of internal ID searching algorithm to avoid O(n) behavior for many common cases. QAK - 2003/10/06 - Allow partial parallel writing to compact datasets. QAK - 2003/10/06 - Correctly create reference to shared datatype in attribute, instead of making a copy of the shared datatype in the attribute. QAK - 2003/10/01 - Revert changes which caused files >2GB to fail when created with MPI-I/O file driver on certain platforms. QAK - 2003/09/16 - Allow compound datatypes to grow in size. SLU - 2003/09/10 - Detect if a type is already packed before attempting to pack it again or check if it is locked. SLU - 2003/09/10 - Corrected bug when opening a file twice with read-only permission for one open and then closing the read-only access file ID would generate an error. QAK - 2003/09/10 - Corrected bug in repeated calls to H5Pget_access_plist() which would incorrectly manage reference counts of internal information and eventually blow up. QAK - 2003/09/02 - Return rank of the array datatype on successful call to H5Tget_array_dims(). QAK - 2003/08/30 - Corrected bug in H5Tdetect_class which was not correctly detecting datatype classes of fields in nested compound datatypes in some circumstances. QAK - 2003/08/30 - Corrected bug in sieve buffer code which could cause loss of data when a small dataset was created and deleted in quick succession. QAK - 2003/08/27 - Corrected bug in H5Gget_objname_by_idx which was not allowing NULL for the name when just querying for the object name's length. QAK - 2003/08/25 - Corrected bug in variable-length string handling which could generate a core dump on writing variable-length strings as part of a compound datatype on certain architectures. QAK - 2003/08/25 - Corrected bug in H5Tget_native_type which would incorrectly compute the size of certain compound datatypes and also incorrectly compute the offset of the last field for those compound datatypes. QAK - 2003/08/25 - Corrected bug in H5Tget_native_type which would drop string datatype metadata (padding, etc.) QAK - 2003/08/25 - Corrected bugs in H5Gget_num_objs, H5Gget_objname_by_idx and H5Gget_objtype_by_idx to allow them to accept location IDs, not just group IDs. QAK - 2003/08/21 - Corrected bug when using scalar dataspace for memory selection and operating on chunked dataset. QAK - 2003/08/18 - Corrected bugs with multiple '/' characters in names for H5Glink and H5Gunlink. QAK - 2003/08/16 - Corrected bug with user blocks that didn't allow a user block to be inserted in front of a file after the file was created. QAK - 2003/08/13 - Corrected errors with using point selections to access data in chunked datasets. QAK - 2003/07/23 - Corrected error with variable-length datatypes and chunked datasets caused H5Dwrite to fail sometimes. QAK - 2003/07/19 - Modified library and file format to support storing indexed storage (chunked dataset) B-tree's with non-default internal 'K' values. QAK - 2003/07/15 - Returned H5T_BKG_TEMP support to library after it was accidentally removed. QAK - 2003/07/14 Configuration ------------- - Parallel I/O with the MPI-I/O driver will no longer work if the filesystem is not POSIX compliant. The "HDF5_MPI_1_METAWRITE" environment variable has been removed. QAK - 2004/01/30 - Fixed the error that cause "make install" to fail because of the macro definition syntax of "prefix?=..." AKC - 2003/07/22 Performance ------------- - Optimized I/O for enumerated datatypes that are a superset of source enumerated datatype. QAK - 2005/03/19 - More optimizations to inner loops of datatype conversions for integers and floats which give a 10-50% speedup. QAK - 2003/11/07 - Hoisted invariant 'if/else's out of inner datatype conversion loop for integer and floating-point values, giving about a 20% speedup. QAK - 2003/10/20 Tools ----- - Fixed h5dump to print attributes data in ASCII if -r option is used. AKC - 2004/11/18 - Fixed space utilization reported in h5ls to correct error in formula used. QAK - 2004/10/22 - Fixed h5redeploy which sometimes complain too many argument for the test command. (The complain did not hinder the h5redploy to proceed correctly.) AKC - 2003/11/03 - Fixed a segmentation fault of h5diff when percentage option is used. AKC - 2003/08/27 - Switched away from tools using internal "fixtype" function(s) to use H5Tget_native_type() internally. QAK - 2003/08/25 Documentation ------------- F90 APIs -------- - h5pget_driver_f was returning information that could not be interpreted by fortran application program; fixed. EIP - 2005/04/10 Platforms Tested ================ AIX 5.1 (32 and 64-bit) xlc 6.0.0.2 xlf 8.1.0.3 xlC 6.0.0.4 xlc 5.0.2.5 xlf 7.1.1.2 xlC 5.0.2.5 mpcc_r 5.0.2.5 mpxlf_r 7.1.1.2 poe 3.2.0.10 Cray T3E sn6606 2.0.6.08 Cray Standard C Version 6.6.0.2 Cray Fortran Version 3.6.0.0.2 mpt 2.2.0.0 Cray SV1 sn9617 10.0.1.2 Cray Standard C Version 6.6.0.2 mpt 2.2.0.0 Cray Fortran Version 3.6.0.0.2 Cray T90IEEE 10.0.1.01y Cray Standard C Version 6.4.0.2.3 Cray Fortran Version 3.4.0.3 mpt 2.1.0.0 FreeBSD 4.9 gcc 2.95.4 g++ 2.95.4 HP-UX B.11.00 HP C HP92453-01 A.11.01.20 HP F90 v2.4 HP ANSI C++ B3910B A.03.13 MPIch 1.2.4 IRIX 6.5 MIPSpro cc 7.30 IRIX64 6.5 (64 & n32) MIPSpro cc 7.3.1.3m F90 MIPSpro 7.3.1.3m (64 only) MPIch 1.2.4 Linux 2.4.18 gcc 2.96, 3.2.2, 3.2.3 g++ 3.2.2, 3.2.3 Intel(R) C++ Version 7.1 Intel(R) Fortran Compiler Version 7.1 PGI compilers (pgcc, pgf90, pgCC) version 5.0-2 MPIch 1.2.4 Absoft Fortran v9.0 OSF1 V5.1 Compaq C V6.4-014 Compaq C V6.3-027 Compaq Fortran V5.5-1877 Compaq C++ V6.5-014 MPI_64bit_R5 g++ version 3.0 for C++ SunOS 5.7 WorkShop Compilers 5.0 98/12/15 C 5.0 (Solaris 2.7) WorkShop Compilers 5.0 98/12/15 C++ 5.0 WorkShop Compilers 5.0 98/10/25 FORTRAN 90 2.0 Patch 107356-04 SunOS 5.8 32,46 Sun WorkShop 6 update 2 C 5.3 (Solaris 2.8) Sun WorkShop 6 update 2 Fortran 90 Sun WorkShop 6 update 2 C++ 5.3 SunOS 5.9 32,64 Sun C 5.6 2004/07/15 (Solaris 2.9) Sun Fortran 95 8.0 2004/07/15 Sun C++ 5.6 2004/07/15 TFLOPS r1.0.4 v4.3.3 i386 pgcc Rel 3.1-4i with mpich-1.2.4 with local modifications IA-32 Linux 2.4.9 gcc 2.96 Intel(R) C++ Version 7.0 Intel(R) Fortran Compiler Version 7.0 IA-64 Linux 2.4.16 ia64 gcc version 2.96 20000731 Intel(R) C++ Version 7.0 Intel(R) Fortran Compiler Version 7.0 Windows XP MSVC++ 6.0 MSVC++ .NET 2003, Intel C++ Version 8.1 DEC Visual Fortran 6.6c MAC OS X Darwin 6.5 gcc and g++ Apple Computer, Inc. GCC version 1161, based on gcc version 3.1 IBM XL Fortran compiler version 8.1 Absoft Fortran v8.2 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 dna = does not apply ( ) = footnote appears below second table Platform C C F90 F90 C++ Shared zlib parallel parallel libraries (4) Solaris2.7 64-bit y y (1) y y (1) y y y Solaris2.7 32-bit y y (1) y y (1) y y y Solaris2.8 64-bit y y (1) y y (1) y y y Solaris2.8 32-bit y y y y (1) y y y IRIX6.5 y y (1) n n n y y IRIX64_6.5 64-bit y y (2) y y y y y IRIX64_6.5 32-bit y y (2) n n n y y HPUX11.00 y y (1) y y y y y OSF1 v5.1 y y y y y y y T3E y y (5) y y (5) n n y SV1 y y (5) y y (5) n n y T90 IEEE y y (5) y y (5) n n y TFLOPS n y (1) n n n n y AIX-5.1 32-bit y y y y y n y AIX-5.1 64-bit y y y y y n y WinXP (6) y n n n y y y WinXP Intel y n n n y y y WinNT CW y n n n n n y Mac OS X 10.2 y n n n y y y FreeBSD y y (1) n n y y y Linux 2.4 gcc (3) y y (1) y (PGI) n y y y Linux 2.4 Intel (3) y n y n y n y Linux 2.4 PGI (3) y n y n y n y Linux 2.4 IA32 Intel y n y n y n y Linux 2.4 IA64 Intel y n y n y n y ASCII Table 2 -- for RELEASE.txt Platform static- Thread- SZIP GASS STREAM- High-level H4/H5 exec safe VFD APIs tools (7) Solaris2.7 64-bit x y y n y y n Solaris2.7 32-bit x y y n y y y Solaris2.8 64-bit x y y n y y n Solaris2.8 32-bit x y y n y y y IRIX6.5 x n y n y y y IRIX64_6.5 64-bit x y y y y y y IRIX64_6.5 32-bit x y y y y y y HPUX11.00 x n y n y y y OSF1 v5.1 y n y n y y y T3E y n n n y y y SV1 y n n n y y y T90 IEEE y n n n y y n TFLOPS y n n n n n n AIX-5.1 32-bit y n y n y y y AIX-5.1 64-bit y n y n y y y WinXP (6) y n y n n y y WinXP Intel y n y n n y y WinNT CW y n y n n y y Mac OS X 10.2 y n y n y y n FreeBSD y y y n y y y Linux 2.4 gcc (3) y y y n y y y Linux 2.4 Intel (3) y n y n y n n Linux 2.4 PGI (3) y n y n y n n Linux 2.4 IA32 Intel y n y n y y y Linux 2.4 IA64 Intel y n y n y y y Notes: (1) Using mpich 1.2.4. (2) Using mpt and mpich 1.2.4. (3) Linux 2.4 with GNU, Intel, and PGI compilers, 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) Binaries only; source code for this platform is not being released at this time. (7) Includes the H4toH5 Library and the h4toh5 and h5toh4 utilities. Compiler versions for each platform are listed in the preceding "Platforms Tested" table. Known Problems ============== * 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 windows XP, Fortran DLLs are not working with DEC Fortran 6.6c and Intel 8.1. * 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 setenv CXX "pgCC -tlocal" for others * 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. 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. * 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://hdf.ncsa.uiuc.edu/pub/outgoing/hdf5/hdf5-1.6.2/F90_source_for_Crays * 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) * On IA32 and IA64 systems, if you use a compiler other than GCC (such as Intel's ecc or icc compilers), you will need to modify the generated "libtool" program after configuration is finished. On or around line 104 of the libtool file, there are lines which look like: # How to pass a linker flag through the compiler. wl="" change these lines to this: # How to pass a linker flag through the compiler. wl="-Wl," UPDATE: This is now done automatically by the configure script. However, if you still experience a problem, you may want to check this line in the libtool file and make sure that it has the correct value. * 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.