HDF5 version 1.10.2-snap4 currently under development ================================================================================ INTRODUCTION This document describes the differences between HDF5-1.10.1 and HDF5 1.10.2, and contains information on the platforms tested and known problems in HDF5-1.10.1. For more details check the HISTORY*.txt files in the HDF5 source. Links to HDF5 1.10.1 source code, documentation, and additional materials can be found on The HDF5 web page at: https://support.hdfgroup.org/HDF5/ The HDF5 1.10.1 release can be obtained from: https://support.hdfgroup.org/HDF5/release/obtain5.html User documentation for the snapshot can be accessed directly at this location: https://support.hdfgroup.org/HDF5/doc/ New features in the HDF5-1.10.x release series, including brief general descriptions of some new and modified APIs, are described in the "New Features in HDF5 1.10" document: https://support.hdfgroup.org/HDF5/docNewFeatures/index.html All new and modified APIs are listed in detail in the "HDF5 Software Changes from Release to Release" document, in the section "Release 1.10.1 (current release) versus Release 1.10.0 https://support.hdfgroup.org/HDF5/doc/ADGuide/Changes.html 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 HDF5-1.10.1 - Supported Platforms - Tested Configuration Features Summary - More Tested Platforms - Known Problems New Features ============ Configuration: ------------- - Removed version-specific gcc/gfortran flags for version 4.0 (inclusive) and earlier. The config/gnu-flags file, which is sourced as a part of the configure process, adds version-specific flags for use when building HDF5. Most of these flags control warnings and do not affect the final product. Flags for older versions of the compiler were consolidated into the common flags section. Moving these flags simplifies maintenance of the file. The upshot of this is that building with ancient versions of gcc (<= 4.0) will possibly no longer work without hand-hacking the file to remove the flags not understood by that version of the compiler. Nothing should change when building with gcc >= 4.1. (HDFFV-9937, DER, 2017/05/31) - -fno-omit-frame-pointer was added when building with debugging symbols enabled. Debugging symbols can be enabled independently of the overall build mode in both the autotools and CMake. This allows (limited) debugging of optimized code. Since many debuggers rely on the frame pointer, we've disabled this optimization when debugging symbols are requested (e.g.: via building with --enable-symbols). (HDFFV-10226, DER, 2017/05/31) Library: -------- - H5FDdriver_query() API call added to the C library. This new library call allows the user to query a virtual file driver (VFD) for the feature flags it supports (listed in H5FDpublic.h). This can be useful to determine if a VFD supports SWMR, for example. Note that some VFDs have feature flags that may only be present after a file has been created or opened (e.g.: the core VFD will have the H5FD_FEAT_POSIX_COMPAT_HANDLE flag set if the backing store is switched on). Since the new API call queries a generic VFD unassociated with a file, these flags will never be returned. (HDFFV-10215, DER, 2017/05/31) - H5FD_FEAT_DEFAULT_VFD_COMPATIBLE VFD feature flag added to the C library. This new feature flag indicates that the VFD is compatible with the default VFD. VFDs that set this flag create single files that follow the canonical HDF5 file format. (HDFFV-10214, DER, 2017/05/31) - The H5I_REFERENCE value in the H5I_type_t enum (defined in H5Ipublic.h) has been marked as deprectated. This ID type value is not used in the C library. i.e.: There are no hid_t values that are of ID type H5I_REFERENCE. This enum value will be removed in a future major version of the library. The code will remain unchanged in the HDF5 1.10.x branches. (HDFFV-10252, DER, 2017/04/05) Parallel Library: ----------------- - Optimize parallel open/location of the HDF5 super-block Previous releases of PHDF5 required all parallel ranks to search for the HDF5 superblock signature when opening the file. As this is accomplished more or less as a synchronous operation, a large number of processes can experience a slowdown in the file open due to filesystem contention. As a first step in improving the startup/file-open performance, we allow MPI rank 0 of the associated MPI communicator to locate the base offset of the super-block and then broadcast that result to the remaining ranks in the parallel group. Note that this approach is utilized ONLY during file opens which employ the MPIO file driver in HDF5 by previously having called H5Pset_fapl_mpio(). HDF5 parallel file operations which do not employ multiple ranks e.g. specifiying MPI_COMM_SELF (whose MPI_Comm_size == 1) as opposed to MPI_COMM_WORLD, will not be affected by this optimization. Conversely, parallel file operations on subgroups of MPI_COMM_WORLD are allowed to be run in parallel with each subgroup operating as an independant collection of processes. (RAW - 2017/10/10, HDFFV-10294) - Large MPI-IO transfers Previous releases of PHDF5 would fail when attempting to read or write greater than 2GB of data in a single IO operation. This issue stems principally from an MPI API whose definitions utilize 32 bit integers to describe the number of data elements and datatype that MPI should use to effect a data transfer. Historically, HDF5 has invoked MPI-IO with the number of elements in a contiguous buffer represented as the length of that buffer in bytes. Resolving the issue and thus enabling larger MPI-IO transfers is accomplished first, by detecting when a user IO request would exceed the 2GB limit as described above. Once a transfer request is identified as requiring special handling, PHDF5 now creates a derived datatype consisting of a vector of fixed sized blocks which is in turn wrapped within a single MPI_Type_struct to contain the vector and any remaining data. The newly created datatype is then used in place of MPI_BYTE and can be used to fulfill the original user request without encountering API errors. (RAW - 2017/09/10, HDFFV-8839) Fortran Library: ---------------- - C++ Library: ------------ - Java Library: ------------- - The H5I_REFERENCE value in the H5I_type_t enum (defined in H5Ipublic.h) has been marked as deprectated. JNI code which refers to this value will be removed in a future major version of the library. The code will remain unchanged in the 1.10.x branches. See the C library section, above, for further information. (HDFFV-10252, DER, 2017/04/05) Tools: ------ - h5diff h5diff has new option enable-error-stack. Updated h5diff with the --enable-error-stack argument, which enables the display of the hdf5 error stack. This completes the improvement to the main tools; h5copy, h5diff, h5dump, h5ls and h5repack. (ADB - 2017/08/30, HDFFV-9774) High-Level APIs: --------------- - C Packet Table API ------------------ - Internal header file -------------------- - Documentation ------------- - Support for new platforms, languages and compilers. ======================================= - Bug Fixes since HDF5-1.10.1 release ================================== Library ------- - filter plugin handling in H5PL.c and H5Z.c It was discovered that the dynamic loading process used by filter plugins had issues with library dependencies. CMake build process changed to use LINK INTERFACE keywords, which allowed HDF5 C library to make dependent libraries private. The filter plugin libraries no longer require dependent libraries (such as szip or zlib) to be available. (ADB - 2017/11/16, HDFFV-10328) - Fix rare object header corruption bug In certain cases, such as when converting large attributes to dense storage, an error could occur which would either fail an assertion or cause file corruption. Fixed and added test. (NAF - 2017/11/14, HDFFV-10274) - H5Zfilter_avail in H5Z.c The public function checked for plugins, while the private function did not. Modified H5Zfilter_avail and private function, H5Z_filter_avail. Moved check for plugin from public to private function. Updated H5P__set_filter due to change in H5Z_filter_avail. Updated tests. (ADB - 2017/10/10, HDFFV-10297, HDFFV-10319) - Fix H5HL_offset_into() (1) Fix H5HL_offset_into() to return error when offset exceeds heap data block size. (2) Fix other places in the library that call this routine to detect error routine. (HDFFV-10216, VC, 2017/08/30) - Fixes for paged aggregation Skip test in test/fheap.c when: (1) multi/split drivers and (2) persisting free-space or using paged aggregation strategy (VC, 2017/07/10) - Fixes for paged aggregation Changes made based on RFC review comments: (1) Add maximum value for file space page size (2) Drop check for page end metadata threshold (3) Remove "can_shrink" and "shrink" callbacks for small section class (VC, 2017/06/09) - Fix for infinite loop in H5VM_power2up(). The function H5VM_power2up() returns the next power of 2 for n. When n exceeds 2^63, it overflows and becomes 0 causing the infinite looping. The fix ensures that the function checks for n >= 2^63 and returns 0. (HDFFV-10217, VC, 2017/07/10) - Fix for H5Ocopy doesn't work with open identifiers Changes made so that raw data for dataset objects are copied from cached info when possible instead of flushing objects to file and read them back in again. (HDFFV-7853, VC, 2017/07/05) - Refactored the testpar/t_bigio.c test to include ALARM macros Changed the test to include the ALARM_ON and ALARM_OFF macros which are intended to prevent nightly test hangs that have been observed with this particular parallel test example. The code was also modified to simplify status reporting (only from MPI rank 0) and additional status checking added. (RAW - 2017/11/08, HDFFV-10301) - An uninitialized struct could cause a memory access error when using variable-length or reference types in a compressed, chunked dataset. A struct containing a callback function pointer and a pointer to some associated data was used before initialization. This could cause a memory access error and system crash. This could only occur under unusual conditions when using variable-lenth and reference types in a compressed, chunked dataset. On recent versions of Visual Studio, when built in debug mode, the debug heap will complain and cause a crash if the code in question is executed (this will cause the objcopy test to fail). (DER - 2017/11/21, HDFFV-10330) - Fix for collective metadata writes on file close It was discovered that metadata was being written twice as part of the parallel file close behavior, once independently and once collectively. A fix for this error was included as part of the parallel compression feature but remained undocumented here. (RAW - 2017/12/01, HDFFV-10272) Configuration ------------- - cmake The hdf5 library used shared szip and zlib, which needlessly required applications to link with the same szip and zlib libraries. Changed the target_link_libraries commands to use the static libs. Removed improper link duplication of szip and zlib. Adjusted the link dependencies and the link interface values of the target_link_libraries commands. (ADB - 2017/11/14, HDFFV-10329) - cmake MPI CMake implementation for MPI was problematic and would create incorrect MPI library references in the hdf5 libraries. Reworked the CMake MPI code to properly create CMake targets.Also merged the latest CMake FindMPI.cmake changes to the local copy. This is necessary until HDF changes the CMake minimum to 3.9 or greater. (ADB - 2017/11/02, HDFFV-10321) - Corrected FORTRAN_HAVE_C_LONG_DOUBLE processing in the autotools. A bug in the autotools Fortran processing code always set the FORTRAN_HAVE_C_LONG_DOUBLE variable to be true regardless of whether or not a C long double type was present. This would cause compilation failures on platforms where a C long double type was not available and the Fortran wrappers were being built. (HDFFV-10247, DER, 2017/07/05) - The deprecated --enable-production and --enable-debug configure options failed to emit errors when passed an empty string (e.g.: --enable-debug=""). Due to the way we checked for these options being set, it was possible to avoid the error message and continue configuration if an empty string was passed to the option. Any use of --enable-production or --enable-debug will now halt the configuration step and emit a helpful error message (use --enable-build-mode=debug|production instead). (HDFFV-10248, DER, 2017/07/05) - cmake Too many commands for POST_BUILD step caused command line to be too big on windows. Changed foreach of copy command to use a custom command with the use of the HDFTEST_COPY_FILE macro. (ADB - 2017/07/12, HDFFV-10254) - cmake test execution environment The parallel HDF5 test: 't_pread' assumed the use of autotools and the directory structure associated with that testing approach. Modified the test code to check whether the 'h5jam' utility can be found in the same directory as the test executable (which is preferred directory structure utilized by cmake) and if found will invoke the tool directly rather than utilizing a relative path. (RAW - 2017/11/03, HDFFV-10318) Performance ------------- - Fortran -------- - Fixed compilation errors when using Intel 18 Fortran compilers (MSB - 2017/11/3, HDFFV-10322) Tools ----- - h5repack h5repack failed to copy a dataset with existing filter. Reworked code for h5repack and h5diff code in tools library. Added improved error handling, cleanup of resources and checks of calls. Modified H5Zfilter_avail and private function, H5Z_filter_avail. Moved check for plugin from public to private function. Updated H5P__set_filter due to change in H5Z_filter_avail. Updated tests. Note, h5repack output display has changed to clarify the individual steps of the repack process. The output indicates if an operation applies to all objects. Lines with notation and no information have been removed. (ADB - 2017/10/10, HDFFV-10297, HDFFV-10319) - h5repack h5repack always set the User Defined filter flag to H5Z_FLAG_MANDATORY. Added another parameter to the 'UD=' option to set the flag by default to '0' or H5Z_FLAG_MANDATORY, the other choice is '1' or H5Z_FLAG_OPTIONAL. (ADB - 2017/08/31, HDFFV-10269) - h5ls h5ls generated error on stack when it encountered a H5S_NULL dataspace. Adding checks for H5S_NULL before calling H5Sis_simple (located in the h5tools_dump_mem function) fixed the issue. (ADB - 2017/08/17, HDFFV-10188) - h5repack Add tests to h5repack.sh.in to verify options added for paged aggregation work as expected. (VC, 2017/08/03) - h5dump h5dump segfaulted on output of XML file. Function that escape'd strings used the full buffer length instead of just the length of the replacement string in a strncpy call. Using the correct length fixed the issue. (ADB - 2017/08/01, HDFFV-10256) - h5diff h5diff segfaulted on compare of a NULL variable length string. Improved h5diff compare of strings by adding a check for NULL strings and setting the lengths to zero. (ADB - 2017/07/25, HDFFV-10246) - h5import h5import crashed trying to import data from a subset of a dataset. Improved h5import by adding the SUBSET keyword. h5import understands to use the Count times the Block as the size of the dimensions. Added INPUT_B_ORDER keyword to old-style configuration files. The import from h5dump function expects the binary files to use native types (FILE '-b' option) in the binary file. (ADB - 2017/06/15, HDFFV-10219) - h5repack h5repack did not maintain the creation order flag of the root group. Improved h5repack by reading the creation order and applying the flag to the new root group. Also added arguments to set the order and index direction, which applies to the traversing of the original file, on the command line. (ADB - 2017/05/26, HDFFV-8611) - h5diff h5diff failed to account for strpad type and null terminators of char strings. Also, h5diff failed to account for string length differences and would give a different result depending on file order in the command line. Improved h5diff compare of strings and arrays by adding a check for string lengths and if the strpad was null filled. (ADB - 2017/05/18, HDFFV-9055, HDFFV-10128) High-Level APIs: ------ - Fixed a bug in the H5DOwrite_chunk() API call where overwriting a chunk with no filters caused an assert to trip (with assertions enabled / debug builds) or additional chunks to be inserted instead of overwriting the data (without assertions enabled / production builds). (HDFFV-10187, DER, 2017/05/11) Fortran High-Level APIs: ------ - Documentation ------------- - F90 APIs -------- - C++ APIs -------- - Testing ------- - Supported Platforms =================== Linux 2.6.32-573.18.1.el6.ppc64 gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-4) #1 SMP ppc64 GNU/Linux g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-4) (ostrich) GNU Fortran (GCC) 4.4.7 20120313 (Red Hat 4.4.7-4) IBM XL C/C++ V13.1 IBM XL Fortran V15.1 Linux 3.10.0-327.10.1.el7 GNU C (gcc), Fortran (gfortran), C++ (g++) #1 SMP x86_64 GNU/Linux compilers: (kituo/moohan) Version 4.8.5 20150623 (Red Hat 4.8.5-4) Version 4.9.3, Version 5.2.0 Intel(R) C (icc), C++ (icpc), Fortran (icc) compilers: Version 15.0.3.187 Build 20150407 MPICH 3.1.4 compiled with GCC 4.9.3 SunOS 5.11 32- and 64-bit Sun C 5.12 SunOS_sparc (emu) Sun Fortran 95 8.6 SunOS_sparc Sun C++ 5.12 SunOS_sparc Windows 7 Visual Studio 2012 w/ Intel Fortran 15 (cmake) Visual Studio 2013 w/ Intel Fortran 15 (cmake) Visual Studio 2015 w/ Intel Fortran 16 (cmake) Windows 7 x64 Visual Studio 2012 w/ Intel Fortran 15 (cmake) Visual Studio 2013 w/ Intel Fortran 15 (cmake) Visual Studio 2015 w/ Intel Fortran 16 (cmake) Visual Studio 2015 w/ Intel C, Fortran 2017 (cmake) Visual Studio 2015 w/ MSMPI 8 (cmake) Cygwin(CYGWIN_NT-6.1 2.8.0(0.309/5/3) gcc and gfortran compilers (GCC 5.4.0) (cmake and autotools) Windows 10 Visual Studio 2015 w/ Intel Fortran 16 (cmake) Cygwin(CYGWIN_NT-6.1 2.8.0(0.309/5/3) gcc and gfortran compilers (GCC 5.4.0) (cmake and autotools) Windows 10 x64 Visual Studio 2015 w/ Intel Fortran 16 (cmake) Mac OS X Mt. Lion 10.8.5 Apple clang/clang++ version 5.1 from Xcode 5.1 64-bit gfortran GNU Fortran (GCC) 4.8.2 (swallow/kite) Intel icc/icpc/ifort version 15.0.3 Mac OS X Mavericks 10.9.5 Apple clang/clang++ version 6.0 from Xcode 6.2 64-bit gfortran GNU Fortran (GCC) 4.9.2 (wren/quail) Intel icc/icpc/ifort version 15.0.3 Mac OS X Yosemite 10.10.5 Apple clang/clang++ version 6.1 from Xcode 7.0 64-bit gfortran GNU Fortran (GCC) 4.9.2 (osx1010dev/osx1010test) Intel icc/icpc/ifort version 15.0.3 Mac OS X El Capitan 10.11.6 Apple clang/clang++ version 7.3.0 from Xcode 7.3 64-bit gfortran GNU Fortran (GCC) 5.2.0 (osx1010dev/osx1010test) Intel icc/icpc/ifort version 16.0.2 Tested Configuration Features Summary ===================================== In the tables below y = tested n = not tested in this release C = Cluster W = Workstation x = not working in this release dna = does not apply ( ) = footnote appears below second table = testing incomplete on this feature or platform Platform C F90/ F90 C++ zlib SZIP parallel F2003 parallel Solaris2.11 32-bit n y/y n y y y Solaris2.11 64-bit n y/n n y y y Windows 7 y y/y n y y y Windows 7 x64 y y/y y y y y Windows 7 Cygwin n y/n n y y y Windows 7 x64 Cygwin n y/n n y y y Windows 10 y y/y n y y y Windows 10 x64 y y/y n y y y Mac OS X Mountain Lion 10.8.5 64-bit n y/y n y y y Mac OS X Mavericks 10.9.5 64-bit n y/y n y y y Mac OS X Yosemite 10.10.5 64-bit n y/y n y y y Mac OS X El Capitan 10.11.6 64-bit n y/y n y y y CentOS 7.2 Linux 2.6.32 x86_64 PGI n y/y n y y y CentOS 7.2 Linux 2.6.32 x86_64 GNU y y/y y y y y CentOS 7.2 Linux 2.6.32 x86_64 Intel n y/y n y y y Linux 2.6.32-573.18.1.el6.ppc64 n y/y n y y y Platform Shared Shared Shared Thread- C libs F90 libs C++ libs safe Solaris2.11 32-bit y y y y Solaris2.11 64-bit y y y y Windows 7 y y y y Windows 7 x64 y y y y Windows 7 Cygwin n n n y Windows 7 x64 Cygwin n n n y Windows 10 y y y y Windows 10 x64 y y y y Mac OS X Mountain Lion 10.8.5 64-bit y n y y Mac OS X Mavericks 10.9.5 64-bit y n y y Mac OS X Yosemite 10.10.5 64-bit y n y y Mac OS X El Capitan 10.11.6 64-bit y n y y CentOS 7.2 Linux 2.6.32 x86_64 PGI y y y n CentOS 7.2 Linux 2.6.32 x86_64 GNU y y y y CentOS 7.2 Linux 2.6.32 x86_64 Intel y y y n Linux 2.6.32-573.18.1.el6.ppc64 y y y n Compiler versions for each platform are listed in the preceding "Supported Platforms" table. More Tested Platforms ===================== The following platforms are not supported but have been tested for this release. Linux 2.6.32-573.22.1.el6 GNU C (gcc), Fortran (gfortran), C++ (g++) #1 SMP x86_64 GNU/Linux compilers: (mayll/platypus) Version 4.4.7 20120313 Version 4.8.4 PGI C, Fortran, C++ for 64-bit target on x86-64; Version 16.10-0 Intel(R) C (icc), C++ (icpc), Fortran (icc) compilers: Version 15.0.3.187 (Build 20150407) MPICH 3.1.4 compiled with GCC 4.9.3 Linux 3.10.0-327.18.2.el7 GNU C (gcc) and C++ (g++) compilers #1 SMP x86_64 GNU/Linux Version 4.8.5 20150623 (Red Hat 4.8.5-4) (jelly) with NAG Fortran Compiler Release 6.1(Tozai) Intel(R) C (icc) and C++ (icpc) compilers Version 15.0.3.187 (Build 20150407) with NAG Fortran Compiler Release 6.1(Tozai) Linux 2.6.32-573.18.1.el6.ppc64 MPICH mpich 3.1.4 compiled with #1 SMP ppc64 GNU/Linux IBM XL C/C++ for Linux, V13.1 (ostrich) and IBM XL Fortran for Linux, V15.1 Debian 8.4 3.16.0-4-amd64 #1 SMP Debian 3.16.36-1 x86_64 GNU/Linux gcc, g++ (Debian 4.9.2-10) 4.9.2 GNU Fortran (Debian 4.9.2-10) 4.9.2 (cmake and autotools) Fedora 24 4.7.2-201.fc24.x86_64 #1 SMP x86_64 x86_64 x86_64 GNU/Linux gcc, g++ (GCC) 6.1.1 20160621 (Red Hat 6.1.1-3) GNU Fortran (GCC) 6.1.1 20160621 (Red Hat 6.1.1-3) (cmake and autotools) Ubuntu 16.04.1 4.4.0-38-generic #57-Ubuntu SMP x86_64 GNU/Linux gcc, g++ (Ubuntu 5.4.0-6ubuntu1~16.04.2) 5.4.0 20160609 GNU Fortran (Ubuntu 5.4.0-6ubuntu1~16.04.2) 5.4.0 20160609 (cmake and autotools) Known Problems ============== At present, metadata cache images may not be generated by parallel applications. Parallel applications can read files with metadata cache images, but since this is a collective operation, a deadlock is possible if one or more processes do not participate. Known problems in previous releases can be found in the HISTORY*.txt files in the HDF5 source. Please report any new problems found to help@hdfgroup.org.