[svn-r29345] Merge of r29326, 29340, 29341 from trunk.
C++ and h5check_version changes.
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ C++ and Fortran
[svn-r29343] Merge of r29332, 29336, 29337 from trunk.
C++ and Fortran changes.
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ C++ and Fortran
[svn-r29335] Merge of r29292, 29301, 29306, 29307, 29312, 29314, 29322, 29325
from trunk.
C++ and a few misc changes.
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ C++ and Fortran
bin/cmakehdf5 w/ C++ and Fortran
[svn-r29328] Change H5D_BT2_NODE_SIZE from 512 to 2048 for performance issue.
Tested on quail, osx1010test, platypus, ostrich, kite, emu, mayll, kituo.
[svn-r29287] Merge of r29269, 29270, 29272 from trunk.
packet table improvements
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ Fortran and C++
[svn-r29286] Merge of r29155, 29162, 29163, 29165, 29169, 29172 from trunk.
Fortran fixes and improvements
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ Fortran and C++
autotools parallel (MPICH 3.1.4) w/ Fortran
[svn-r29264] Checkin of fix for assertion failure exposed by Mark Miller's testhdf5
benchmark.
In essence, the problem was failure to set appropriate tags when protecting
entries in the fractal heap sanity check code. This resulted in H5AC protect
failing when H5C_DO_TAGGING_SANITY_CHECKS is set.
This bug was only detected in Mark Miller's bench mark as the existing test
code did not generate the cache contents necessary to execute the pathway in
which the failure occurred.
Cleaned up t...
[svn-r29253] Merge of r29057 from trunk (dxpl checks in debug builds)
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ Fortran and C++
autotools parallel (MPICH 3.1.4) w/ Fortran
bin/cmakehdf5 (CMake 3.3.2)
[svn-r29224] Brought alpha2 branch in sync with revise_chunks (up to r29219)
Note: generated autotools files are from a VM and not THG
machines. These will be updated immediately after
check-in.
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial w/ Fortran & C++
autotools parallel (MPICH 3.1.4) w/ Fortran
CMake serial (cmakehdf5 w/ CMake 3.3.2)
[svn-r29210] Merge of r28950 and 28957 from trunk
Tested on: 64-bit Ubuntu 15.10 (Linux 4.2.0 x86_64) gcc 5.2.1
autotools serial w/ Fortran & C++
autotools parallel (MPICH 3.1.4) w/ Fortran
[svn-r29193] Added POSIX wrappers for a few SWMR test functions that
were missed.
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1
autotools serial
CMake serial (build only)
[svn-r29189] Updated CMake to build SWMR and SWMR VDS test files.
Many SWMR test files were updated to include h5test.h and use HD posix function wrappers so they will compile on Windows.
Tested on: 64-bit Ubuntu 15.10 w/ gcc 5.2.1 & CMake 3.3.2
64-bit Windows 10 w/ VS 2015 Pro & CMake 3.3.1
[svn-r29159] Merge of r29019 from trunk.
Fixes double declaration of a couple of types from the LFS changes.
Tested on: 64-bit Windows 10 w/ VS 2015 Pro
[svn-r29136] Added code to main routine of objcopy to close the FAPL fapl2
shortly before exit.
This in turn hides the property list management failure that
is otherwise expressed on library shutdown when run with either
the split or multi file drivers.
Needless to say, this is a bandaid. The actual cause of the
problem must be found and dealt with.
Tested on Jam -- serial/debug/check-vfd
[svn-r29114] Updated test code of HDFFV-9672 to correct error exposed by
builds with the --with-default-api-version=v16 configure option.
Tested on Mercury (serial/debug & serial/debug/v16_api)
Jam (serial/debug)
NOTE: In the serial/debug test on Mercury, I encountered the
following assertion failure in testh5watch.sh:
H5Fsfile.c:63: H5F_sfile_assert_num: Assertion `H5F_sfile_head_g == ((void *)0)' failed.
As this was clearly unrelated to my change, I went ahe...
[svn-r29054] Description:
Correct error computing the number of bytes to encode dimensions with when
using "latest format" by including the datatype's size in the computation.
Tested on:
Mac OSX/64 10.11.3 (amazon) w/serial & parallel
(h5committest not required on this branch)
[svn-r29052] Merged r28735 from HDF5 1.10.0 alpha branch.
Fix for incorrect chunk size setup with array types in new
chunk indices (extensible and fixed arrays, v2 B-trees)
which was reported by ITER (JIRA issue SWMR-99).
Tested on: 64-bit Ubuntu 15.10 (Linux 4.2.0 x86_64) gcc 5.2.1
Autotools serial, parallel (MPICH 3.1.4)
[svn-r29048] Fix to h5format_convert tool:
(1) Downgrade layout version to 3 for contiguous and compact dataset
(2) Add new test files for testing with/without messages in superblock extension.
Tested on jam, osx1010test, platypus, quail, ostrich, kite, moohan, kituo, mayll, emu.
[svn-r29028] Description:
When SWMR writes are enabled, correct record update algorithm to
indicate to parent node when a child node that was "only" modified was also
"shadowed", so that parent node can update itself (since the address of its
child node will move when it is shadowed).
Re-enable v2 B-tree SWMR testing.
Tested on:
MacOSX/64 10.11.3 (amazon) w/serial & parallel
(h5committest not required on this branch)
[svn-r29016] Merge of r28936, 28949, 28960-29009 from trunk.
Tested on: 64-bit Ubuntu 15.10 (Linux 4.2.0 x86_64) gcc 5.2.1
autotools serial w/ Fortran and C++
autotools parallel (MPICH 3.1.4) w/ Fortran
CMake serial (3.3.2, via cmakehdf5)
[svn-r29014] h5format_convert(): fix to downgrade superblock version and persistent file space so
1.8 library can open/read the file.
Tested on ostrich, kituo, platypus, moohan, quail, emu.
More tests will be added later.
[svn-r28988] Merge of r28986 from trunk (H5_DLLVAR H5AC_dxpl_id).
Tested on: 64-bit Ubuntu 15.10 (Linux 4.2.0 x86_64) gcc 5.2.1
autotools serial and parallel