HDF5 version 1.8.8-snap7 currently under development ================================================================================ INTRODUCTION ============ This document describes the differences between HDF5-1.8.7 and HDF5 1.8.8-*, and contains information on the platforms tested and known problems in HDF5-1.8.8-*. For more details, see the files HISTORY-1_0-1_8_0_rc3.txt and HISTORY-1_8.txt in the release_docs/ directory of the HDF5 source. Links to the HDF5 1.8.8 source code, documentation, and additional materials can be found on the HDF5 web page at: http://www.hdfgroup.org/products/hdf5/ The HDF5 1.8.8 release can be obtained from: http://www.hdfgroup.org/HDF5/release/obtain5.html User documentation for 1.8.8 can be accessed directly at this location: http://www.hdfgroup.org/HDF5/doc/ New features in the HDF5-1.8.x release series, including brief general descriptions of some new and modified APIs, are described in the "What's New in 1.8.0?" document: http://www.hdfgroup.org/HDF5/doc/ADGuide/WhatsNew180.html All new and modified APIs are listed in detail in the "HDF5 Software Changes from Release to Release" document, in the section "Release 1.8.8 (current release) versus Release 1.8.7": http://www.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, Languages, and Compilers - Bug Fixes since HDF5-1.8.7 - Platforms Tested - Supported Configuration Features Summary - Known Problems New Features ============ Configuration ------------- - None Library ------- - None Parallel Library ---------------- - The library added two new dataset transfer property to control whether to fill 0xff in the destination data or convert to the destination data when overflow happens to ENUM data (Issue 7579). The two new functions are H5Pset(get)_enum_conv_overflow. (SLU - 2011/05/27) - H5Tcreate now supports string type (fixed-length and variable-length). (SLU - 2011/05/20) Tools ----- - None High-Level APIs --------------- - None F90 API ------- - None C++ API ------- - None Support for New Platforms, Languages, and Compilers =================================================== - None Bug Fixes since HDF5-1.8.7 ========================== Configuration ------------- - The --enable-h5dump-packed-bits configure option has been removed, as the h5dump code that this option conditionally enabled is now always compiled into h5dump. Please refer to the h5dump reference manual for usage of the packed bits feature. (MAM - 2011/06/23) - Configure now uses the same flags and symbols in its tests that are used to build the library. - On linux we now link to the bsd_compat library, as per the gcc manual, since we define _BSD_SOURCE. - h5repack: h5repack failed to copy dataset if the layout is changed from chunked with unlimited dims to contiguous. (PC -- 2011/07/15) - h5diff: "--delta" option considers two NaN of the same type are different, which is wrong based on http://www.hdfgroup.org/HDF5/doc/RM/Tools.html#Tools-Diff. (PC -- 2011/07/15) Library ------- - Corrected internal error in library where objects that use named datatypes and were accessed from two different file IDs could confuse the two and cause erroneous failures. This addresses Jira issue #7638. (QAK - 2011/07/18) - In v1.6 library, there was EOA for the whole MULTI file saved in the super block. We took it out in v1.8 library because it's meaningless for the MULTI file. v1.8 library saves the EOA for the metadata file, instead. But this caused some backward compatibility problem. v1.8 library couldn't open the file created with v1.6 library. We fixed the problem by checking the EOA value to detect the file created with v1.6 library. (SLU - 2011/6/22) - When a dataset had filters and reading data failed, the error message didn't say which filter isn't registered. It's fixed now. (SLU - 2011/6/3) Parallel Library ---------------- - None Tools ----- - Fixed h5diff to display instructive error message and exit with 1 when mutually exclusive options (-d, -p and --use-system-epsilon) are used together. HDFFV-7600 (JKM 07/07/2011) - Fixed h5dump to display the first line of each element into correct position for multiple dimention array type. Before this fix, the first line of each element in array were displayed after the last line of previous element without moving to the next line (+indentation). Bug #HDFFV-5878 (JKM 06/15/2011) - Fixed h5dump to display correct value for H5T_STD_I8LE dataset on a system (ppc64, linux, Big-Endian, clustering). Bug #HDFFV-7594 (ABERT & JKM 05/12/2011) - Fixed h5diff to compare file itself correctly. Previously h5diff reported either different or not compatible in certain cases even comparing file itself. This fix also improve performance when comparing same target objects through verifying the obj&file addresses before comparing the details in the objects (ex: datasets or attributes) Bug #HDFFV-5928 (XCAO & JKM 05/06/2011) F90 API ------ - None C++ API ------ - None High-Level APIs: ------ - None Fortran High-Level APIs: ------ - None Platforms Tested ================ The following platforms and compilers have been tested for this release. FreeBSD 8.2-STABLE i386 gcc 4.2.1 [FreeBSD] 20070719 (loyalty) g++ 4.2.1 [FreeBSD] 20070719 gcc 4.6.1 20110422 g++ 4.6.1 20110422 gfortran 4.6.1 20110422 FreeBSD 8.2-STABLE amd64 gcc 4.2.1 [FreeBSD] 20070719 (freedom) g++ 4.2.1 [FreeBSD] 20070719 gcc 4.6.1 20110422 g++ 4.6.1 20110422 gfortran 4.6.1 20110422 Linux 2.6.18-194.3.1.el5PAE gcc (GCC) 4.1.2 and 4.4.2 #1 SMP i686 i686 i386 GNU Fortran (GCC) 4.1.2 20080704 (jam) (Red Hat 4.1.2-48) and 4.4.2 PGI C, Fortran, C++ 10.4-0 32-bit PGI C, Fortran, C++ 10.6-0 32-bit Intel(R) C Compiler for 32-bit applications, Version 11.1 Intel(R) C++ Compiler for 32-bit applications, Version 11.1 Intel(R) Fortran Compiler for 32-bit applications, Version 11.1 MPICH mpich2-1.3.1 compiled with gcc 4.1.2 and gfortran 4.1.2 Linux 2.6.18-238.12.1.el5 gcc 4.1.2 and 4.4.2 #1 SMP x86_64 GNU/Linux GNU Fortran (GCC) 4.1.2 20080704 (koala) (Red Hat 4.1.2-46) and 4.4.2 tested for both 32- and 64-bit binaries Intel(R) C, C++, Fortran Compilers for applications running on Intel(R) 64, Version 11.1. PGI C, Fortran, C++ Version 9.0-4 for 64-bit target on x86-64 MPICH mpich2-1.3.1 compiled with gcc 4.1.2 and gfortran 4.1.2 SGI ProPack 7 Linux Intel(R) C++ Version 11.1 20100806 2.6.32.19-0.3.1.1982.0.PTF- Intel(R) Fortran Version 11.1 20100806 default #1 SMP SGI MPT 2.01 SGI Altix UV (NCSA ember) SunOS 5.10 32- and 64-bit Sun C 5.9 Sun OS_sparc Patch 124867-16 (linew) Sun Fortran 95 8.3 Sun OS_sparc Patch 127000-13 Sun C++ 5.9 Sun OS_sparc Patch 124863-62 Sun C 5.10 SunOS_sparc Patch 141861-07 Sun Fortran 95 8.4 SunOS_sparc Patch 128231-06 Sun C++ 5.10 SunOS_sparc 128228-11 Windows XP Visual Studio 2008 w/ Intel Fortran 10.1 (project files) Visual Studio 2008 w/ Intel Fortran 11.1 (cmake) Visual Studio 2010 (cmake) Cygwin(1.7.7 native gcc(4.3.4) compiler and gfortran) Windows XP x64 Visual Studio 2008 w/ Intel Fortran 10.1 (project files) Visual Studio 2008 w/ Intel Fortran 11.1 (cmake) Visual Studio 2010 (cmake) Cygwin(1.7.7 native gcc(4.3.4) compiler and gfortran) Windows Vista Visual Studio 2008 w/ Intel Fortran 11.1 (cmake) Windows Vista x64 Visual Studio 2008 w/ Intel Fortran 11.1 (cmake) Windows 7 Visual Studio 2008 w/ Intel Fortran 11.1 (cmake) Windows 7 x64 Visual Studio 2008 w/ Intel Fortran 11.1 (cmake) Mac OS X 10.6.3 (Intel 64-bit) i686-apple-darwin10-gcc-4.2.1 (GCC) 4.2.1 Darwin Kernel Version 10.3.1 GNU Fortran (GCC) 4.5.0 20090910 Intel C, C++ and Fortran compilers 11.1 20100806 Mac OS X 10.6.4 (Intel 32-bit) i686-apple-darwin10-gcc-4.2.1 (GCC) 4.2.1 Darwin Kernel Version 10.4.0 GNU Fortran (GCC) 4.6.0 20101106 Intel C, C++ and Fortran compilers 12.0.0 20101110 Mac OS X 10.6.4 (Intel 64-bit) i686-apple-darwin10-gcc-4.2.1 (GCC) 4.2.1 (Apple Inc. build 5659) Darwin Kernel Version 10.6.0 GNU Fortran (GCC) 4.5.0 20090910 Intel C, C++ and Fortran compilers 11.1 20100806 Fedora 12 2.6.32.16-150.fc12.ppc64 #1 SMP ppc64 GNU/Linux gcc (GCC) 4.4.4 20100630 (Red Hat 4.4.4-10) GNU Fortran (GCC) 4.4.4 20100630 (Red Hat 4.4.4-10) Debian6.01 2.6.32-5-686 #1 SMP i686 GNU/Linux gcc (Debian 4.4.5-8) 4.4.5 GNU Fortran (Debian 4.4.5-8) 4.4.5 Debian6.01 2.6.32-5-amd64 #1 SMP x86_64 GNU/Linux gcc (Debian 4.4.5-8) 4.4.5 GNU Fortran (Debian 4.4.5-8) 4.4.5 Fedora14 2.6.35.12-88.fc14.i686.PAE #1 SMP i686 i686 i386 GNU/Linux gcc (GCC) 4.5.1 20100924 (Red Hat 4.5.1-4) GNU Fortran (GCC) 4.5.1 20100924 (Red Hat 4.5.1-4) Fedora14 2.6.35.12-88.fc14.x86_64 #1 SMP x86_64 x86_64 x86_64 GNU/Linux gcc (GCC) 4.5.1 20100924 (Red Hat 4.5.1-4) GNU Fortran (GCC) 4.5.1 20100924 (Red Hat 4.5.1-4) SUSE 11.4 2.6.37.1-1.2-desktop #1 SMP PREEMPT i686 i686 i386 GNU/Linux gcc (SUSE Linux) 4.5.1 20101208 GNU Fortran (SUSE Linux) 4.5.1 20101208 SUSE 11.4 2.6.37.1-1.2-desktop #1 SMP PREEMPT x86_64 x86_64 x86_64 GNU/Linux gcc (SUSE Linux) 4.5.1 20101208 GNU Fortran (SUSE Linux) 4.5.1 20101208 Ubuntu 10.10 2.6.35-28-generic #50-Ubuntu SMP i686 GNU/Linux gcc (Ubuntu/Linaro 4.4.4-14ubuntu5) 4.4.5 GNU Fortran (Ubuntu/Linaro 4.4.4-14ubuntu5) 4.4.5 Ubuntu 10.10 2.6.35-28-generic #50-Ubuntu SMP x86_64 GNU/Linux gcc (Ubuntu/Linaro 4.4.4-14ubuntu5) 4.4.5 GNU Fortran (Ubuntu/Linaro 4.4.4-14ubuntu5) 4.4.5 OpenVMS Alpha 8.3 HP C V7.3-009 HP Fortran V8.2-104679-48H9K HP C++ V7.3-009 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 parallel Solaris2.10 32-bit n y n y y y Solaris2.10 64-bit n y n y y y Windows XP n y(4) n y y y Windows XP x64 n y(4) n y y y Windows Vista n y(4) n y y y Windows Vista x64 n y(4) n y y y OpenVMS Alpha n y n y y n Mac OS X 10.6 Intel n y n y y y FreeBSD 8.2-STABLE 32- and 64-bit n x n x y y CentOS 5.5 Linux 2.6.18-194 i686 GNU (1)W y y(2) y y y y CentOS 5.5 Linux 2.6.18-194 i686 Intel W n y n y y n CentOS 5.5 Linux 2.6.18-194 i686 PGI W n y n y y n CentOS 5.5 Linux 2.6.16 x86_64 GNU (1) W y y(3) y y y y CentOS 5.5 Linux 2.6.16 x86_64 Intel W n y n y y n CentOS 5.5 Linux 2.6.16 x86_64 PGI W n y n y y y RedHat EL4 2.6.18 Xeon Lustre C y y y y y n Fedora 12 Linux 2.6.32.16-150.fc12.ppc64 n y n y y y SGI Linux 2.6.32.19 y y y y y y Platform Shared Shared Shared Thread- C libs F90 libs C++ libs safe Solaris2.10 32-bit y y y y Solaris2.10 64-bit y y y y Windows XP y y(4) y n Windows XP x64 y y(4) y n Windows Vista y y(4) y y Windows Vista x64 y y(4) y y OpenVMS Alpha n n n n Mac OS X 10.6 y(5) n y n FreeBSD 8.2-STABLE 32- and 64-bit y x x y CentOS 5.5 Linux 2.6.18-128 i686 GNU (1)W y y(2) y y CentOS 5.5 Linux 2.6.18-128 i686 Intel W y y y n CentOS 5.5 Linux 2.6.18-128 i686 PGI W y y y n CentOS 5.5 Linux 2.6.16 x86_64 GNU (1) W y y y y CentOS 5.5 Linux 2.6.16 x86_64 Intel W y y y n CentOS 5.5 Linux 2.6.16 x86_64 PGI W y y y n RedHat EL4 2.6.18 Xeon Lustre C y y y n Fedora 12 Linux 2.6.32.16-150.fc12.ppc64 y y y y SGI Linux 2.6.32.19 y y y y (1) Fortran compiled with gfortran. (2) With PGI and Absoft compilers. (3) With PGI compiler for Fortran. (4) Using Visual Studio 2008 w/ Intel Fortran 10.1 (Cygwin shared libraries are not supported) (5) C and C++ shared libraries will not be built when Fortran is enabled. Compiler versions for each platform are listed in the preceding "Platforms Tested" table. Known Problems ============== * After the shared library support is fixed for some bugs, it is discovered that "make prefix=XXX install" no longer work for shared libraries. It still works correctly for static libraries. Therefore, if you want to install the HDF5 shared libraries in location such as /usr/local/hdf5, you need to specify it via the --prefix option during configure time. E.g, ./configure --prefix=/usr/local/hdf5 ... (AKC - 2011/05/07 HDFFV-7583) * The parallel test, t_shapesame, in testpar/, may run for a long time and may get terminated by the alarm signal. If that happens, one can increase the alarm seconds (default is 1200 seconds = 20 minutes) by setting the environment variable, $HDF5_ALARM_SECONDS, to a larger value such as 3600 which means 60 minutes = 1 hour. Beware that the t_shapesame test may fail in some systems (see the "While working on the 1.8.6 release..." problem below). If it does, it will waste more time if $HDF5_ALARM_SECONDS is set to a larger value. (AKC - 2011/05/07) * Shared Fortran libraries are not quite working on AIX. While they are generated when --enable-shared is specified, the fortran and hl/fortran tests fail. We are looking into the issue. HL and C++ shared libraries should now be working as intended, however. (MAM - 2011/04/20) * The --with-mpe configure option does not work with Mpich2. AKC - 2011/03/10 * If parallel gmake (e.g., gmake -j 4) is used, the "gmake clean" command sometimes fails in the perform directory due to the attempt to remove the executable of h5perf or h5perf_serial by two "parallel" commands. This error has no consequence on the functionality of the HDF5 library or install. It is fixed in the next release. AKC - 2011/01/25 ??????? 4.18.2011. sent question to albert re: has this been fixed in 1.8.7????????????????????????? * While working on the 1.8.6 release of HDF5, a bug was discovered that can occur when reading from a dataset in parallel shortly after it has been written to collectively. The issue was exposed by a new test in the parallel HDF5 test suite, but had existed before that. We believe the problem lies with certain MPI implementations and/or file systems. We have provided a pure MPI test program, as well as a standalone HDF5 program, that can be used to determine if this is an issue on your system. They should be run across multiple nodes with a varying number of processes. These programs can be found at: http://www.hdfgroup.org/ftp/HDF5/examples/known_problems/ * The library's test dt_arith.c showed a compiler's rounding problem on Cygwin when converting from unsigned long long to long double. The library's own conversion works fine. We defined a macro for Cygwin to skip this test until we can solve the problem. Please see bug #1813. SLU - 2010/5/5 * All the VFL drivers aren't backward compatible. In H5FDpublic.h, the structure H5FD_class_t changed in 1.8. There is new parameter added to get_eoa and set_eoa callback functions. A new callback function get_type_map was added in. The public function H5FDrealloc was taken out in 1.8. The problem only happens when users define their own driver for 1.6 and try to plug in 1.8 library. Because there's only one user complaining about it, we (Elena, Quincey, and I) decided to leave it as it is (see bug report #1279). Quincey will make a plan for 1.10. SLU - 2010/2/2 * MinGW has a missing libstdc++.dll.a library file and will not successfully link C++ applications/tests. Do not use the enable-cxx configure option. Read all of the INSTALL_MINGW.txt file for all restrictions. ADB - 2009/11/11 * The PathScale MPI implementation, accessing a Panasas file system, would cause H5Fcreate() with H5F_ACC_EXCL to fail even when the file does not exist. This is due to the MPI_File_open() call failing if the mode has the MPI_MODE_EXCL bit set. (See bug 1468 for details.) AKC - 2009/8/11 * Parallel tests failed with 16 processes with data inconsistency at testphdf5 / dataset_readAll. Parallel tests also failed with 32 and 64 processes with collective abort of all ranks at t_posix_compliant / allwrite_allread_blocks with MPI IO. CMC - 2009/04/28 * For Red Storm, a Cray XT3 system, the tools/h5ls/testh5ls.sh and tools/h5copy/testh5copy.sh will fail some of its sub-tests. These sub-tests are expected to fail and should exit with a non-zero code but the yod command does not propagate the exit code of the executables. Yod always returns 0 if it can launch the executable. The test suite shell expects a non-zero for this particular test, therefore it concludes the test has failed when it receives 0 from yod. Skip all the "failing" test for now by changing them as following. * The STDIO VFD does not work on some architectures, possibly due to 32/64 bit or large file issues. The basic STDIO VFD test is known to fail on 64-bit SunOS 5.10 on SPARC when built with -m64 and 32-bit OS X/Darwin 10.7.0. The STDIO VFD test has been disabled while we investigate and a fix should appear in 1.8.8. ======== Original tools/h5ls/testh5ls.sh ========= TOOLTEST tgroup-1.ls 1 -w80 -r -g tgroup.h5 ======== Change to =============================== echo SKIP TOOLTEST tgroup-1.ls 1 -w80 -r -g tgroup.h5 ================================================== ======== Original tools/h5copy/testh5copy.sh ========= TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -v -s grp_dsets -d grp_rename TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -v -s grp_dsets -d /grp_rename/grp_dsets TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -vp -s /grp_dsets -d /E/F/grp_dsets TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -vp -s /grp_nested -d /G/H/grp_nested H5LSTEST $FILEOUT ======== Change to =============================== echo SKIP TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -v -s grp_dsets -d grp_rename echo SKIP TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -v -s grp_dsets -d /grp_rename/grp_dsets echo SKIP TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -vp -s /grp_dsets -d /E/F/grp_dsets echo SKIP TOOLTEST_FAIL -i $TESTFILE -o $FILEOUT -vp -s /grp_nested -d /G/H/grp_nested echo SKIP H5LSTEST $FILEOUT ================================================== AKC - 2008/11/10 * For Red Storm, a Cray XT3 system, the yod command sometimes gives the message, "yod allocation delayed for node recovery". This interferes with test suites that do not expect to see this message. See the section of "Red Storm" in file INSTALL_parallel for a way to deal with this problem. AKC - 2008/05/28 * On an Intel 64 Linux cluster (RH 4, Linux 2.6.9) with Intel 10.0 compilers, use -mp -O1 compilation flags to build the libraries. A higher level of optimization causes failures in several HDF5 library tests. * On mpich 1.2.5 and 1.2.6, if more than two processes contribute no IO and the application asks to do collective IO, we have found that when using 4 processors, a simple collective write will sometimes be hung. This can be verified with t_mpi test under testpar. * A dataset created or rewritten with a v1.6.3 library or after cannot be read with the v1.6.2 library or before when the Fletcher32 EDC filter is enabled. There was a bug in the calculation of the Fletcher32 checksum in the library before v1.6.3; the checksum value was not 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 was no longer the same as before on little-endian system. Library releases after 1.6.4 can still read datasets created or rewritten with an HDF5 library of v1.6.2 or before. SLU - 2005/6/30 * 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", but HDF5 does not use sockets. This failure is due to problems with the poe command trying to set up the debug socket. To resolve this problem, check to see whether there are many old /tmp/s.pedb.* files staying around. These are sockets used by the poe command and left behind due to failed commands. First, ask your system administrator to clean them out. Lastly, request IBM to provide a means to run poe without the debug socket. * The --enable-static-exec configure flag will only statically link libraries if the static version of that library is present. If only the shared version of a library exists (i.e., most system libraries on Solaris, AIX, and Mac, for example, only have shared versions), the flag should still result in a successful compilation, but note that the installed executables will not be fully static. Thus, the only guarantee on these systems is that the executable is statically linked with just the HDF5 library. * There is also a configure error on Altix machines that incorrectly reports when a version of Szip without an encoder is being used. * On cobalt, an SGI Altix SMP ia64 system, Intel compiler version 10.1 (which is the default on that system) does not work properly and results in failures during make check (in a static build) and make installcheck (during a shared build). This appears to be a compiler optimization problem. Reducing optimization by setting CFLAGS to -O1 or below resolves the issue. Alternatively, using a newer version of the compiler (11.0) also works as intended. MAM - 2010/06/01