* Macro cleanup and obvious warning fixes in parallel code.
* Moved H5FD comm and info dup/free wrapper code to a new H5mpi.c file
and separated it to deal with each MPI type separately.
Description:
Applied the typo fixes from user's report.
The previous pull request couldn't be merged because it was too old,
and it was too complicated for me to resolve conflicts.
Platform tested:
Linux/64 (jelly) - very minor
* Updated comments in the library to accurately reflect the semantics
of the H5FD_FEAT_POSIX_COMPAT_HANDLE flag.
* Removed the check for the POSIX compatible handle feature flag
when determining if a VFD supports SWMR.
Partial work for HDFFV-10197:
* Added a new H5FD_FEAT_DEFAULT_VFD_COMPATIBLE feature flag that is
set when a VFD creates output that is compatible with the default
VFD. Will be used in the testing but might also be generally useful
to users.
* commit '54957d37f5aa73912763dbb6e308555e863c43f4':
Commit copyright header change for src/H5PLpkg.c which was added after running script to make changes.
Add new files in release_docs to MANIFEST. Cimmit changes to Makefile.in(s) and H5PL.c that resulted from running autogen.sh.
Merge pull request #407 in HDFFV/hdf5 from ~LRKNOX/hdf5_lrk:hdf5_1_10_1 to hdf5_1_10_1
Change copyright headers to replace url referring to file to be removed and replace it with new url for COPYING file.
(https://jira.hdfgroup.org/browse/HDFFV-10055).
Briefly, in H5C_collective_write() in H5Cmpio.c,
the metadata cache attempts to perform a collective
write of metadata cache entries.
This worked fine as long as all processes had at
least one entry to write.
However, when the process has no entries, the
function tries to participate in the collective write
by calling MPI_File_set_view(),
MPI_File_write_all() and then MPI_File_set_view()
again, to match the calls in H5FD_mpio_write().
After pull request 183, the CGNS test benchmark_hdf5
started failing. On investigation, I determined that
the failure occurred in the first call to MPI_File_set_view()
in the "no data to write" path through H5C_collective_write().
Note that pull request 183 did not create the problem,
it only exposed it. The bug can be observed after pull
request 182 if one executes the CGNS progam
src/ptests/benchmark_hdf5 with 90 processes.
The problem appears to have been that the calls to
MPI_File_set_view() in H5C_collective_write() and
H5FD_mpio_write() were using different values for the
info parameter. I patched the problem by adding a
MPI specific VFD call allowing me to get the MPI_Info
used in H5FD_mpio_write() for use in
MPI_File_set_view() calls in H5C_collective_write().
Tested serial & parallel, debug & production on
Jelly.
- remove H5AC_ind_dxpl_id and use only H5AC_dxpl_id everywhere instead.
- remove flush_me_collectively flag from cache entries
- add a collective sanity check (MPI_Barrier) for every HDF5 API routine that could possibly touch the file. This is trigerred when the environment variable H5_COLL_API_SANITY_CHECK is set to a non 0 digit.
tested on BB-8 with serial and parallel.
Refactor property list code to "deep copy" properties in the correct way,
retraining the rest of the library to copy & release things correctly. This
cleans up another batch of memory leaks, etc. within the library.
Tested on:
MacOSX/64 10.10.5 (amazon) w/serial & parallel
Linux/32 2.6.x (jam) w/serial & parallel
(h5committest forthcoming)
Complete revamp of package initialization/shutdown mechanism in the library.
Each package now has a single init/term routine.
This new way should avoid packages being re-initialized during library
shutdown and is also be _much_ more proactive about giving feedback for
resource leaks internal to the library.
Introduces a new "module" header file for packages in the library
(e.g src/H5Fmodule.h) which sets up some necessary package configuration macros
for the FUNC_ENTER/LEAVE macros. (The VFL drivers have their own slightly
modified version of this header, src/H5FDdrvr_module.h)
Also cleaned up a bunch of resources leaks all across the library and tests,
along with addressing many warnings, as I encountered them.
Tested on:
MacOSX/64 10.10.5 (amazon) w/serial & parallel
Linux/64 3.10.x (kituo) w/serial & parallel
Linux/64 2.6.x (ostrich) w/serial
Clean up EOF code within library and add 'mem_type' parameter to 'get_eof'
VFD callback, to avoid various ambiguous situations, particularly with the
multi VFD. (Supports changes for 'avoid_truncate' feature also)
Tested on:
MacOSX/64 10.10.1 (amazon) w/serial & parallel
h5committest forthcoming
Remove all traces of MPI-POSIX VFD and GPFS detection/code.
Remove remaining traces of stream VFD.
Remove testpar/t_posix_compliant test (it's not actually verifying anything).
Clean up H5D__mpio_opt_possible() further.
Moved environment variable that disables MPI collective operations into
MPI-IO VFD (instead of it being in src/H5S.c).
A few other small code cleanups.
Tested on:
Mac OSX/64 10.9.2 (amazon) w/parallel & serial
Clean up warnings, switch library code to use Standard C/POSIX wrapper
macros, remove internal calls to API routines, update checkapi and checkposix
scripts.
Tested on:
Mac OSX/64 10.8.3 (amazon) w/C++ & FORTRAN
Big-Endian Linux/64 (ostrich)
Clean up warnings and bring changes from Coverity branch back to trunk:
r20451:
Fixed coverity 810
r20454:
Purpose: Fix coverity issue 816
Description: Modified various file drivers to always check the return value from
H5P_get_driver_info.
r20455:
Purpose: Cleanup fix for coverity issue 816
Description: Removed redundant check for "fa" in H5FD_core_open.
Tested on:
Mac OSX/64 10.8.2 (amazon) w/debug
(Too minor to require h5committest)
Clean up file driver property handling, along with some misc. warnings.
Tested on:
Mac OSX/64 10.7.4 (amazon) w/debug, gcc 4.7.1, C++, FORTRAN & threadsafe
(h5committest upcoming)
Refactor function name macros and simplify the FUNC_ENTER macros, to clear
away the cruft and prepare for further cleanups.
Tested on:
Mac OSX/64 10.7.3 (amazon) w/debug, production & parallel
