Attempt to ensure that the parallel cache test runs at a reasonable
speed with large numbers of processors.
Description:
In some cases, the number of random locks and unlocks was a multiple of
the MPI rank.
Solution:
Use rank % 4 instead of simply rank.
Platforms tested:
copper
Misc. update:
Add a file that I forgot in my last checkin.
Description:
t_cache.c is the source file for the new parallel metadata cache test.
Solution:
See above.
Platforms tested:
h5committested before the last checkin. Will run another
h5committest shortly, but it shouldn't be necessary as there
are no new changes.
Misc. update:
1) Add parallel test for metadata cache
2) Split serial test for metadata cache into two parts
3) Fix bug in which cache was flushed needlessly when the
cache wasn't full.
4) Performance improvements
5) Update API for parallel cache coherency bug fix.
Description:
See above.
Solution:
See above.
Platforms tested:
h5committest
Misc. update:
bug fix 504
Description:
testpar/t_mpi would hang if $HDF5_NOCLEANUP is set. E.g.,
% env HDF5_NOCLEANUP=yes mpirun -np 3 ./t_mpi
This happened because the environment variables are not exported to all
mpi processes by the mpirun command. So, some attempted to do cleanup
while others don't and some hang waiting for others to act.
Solution:
Instead individual program checking getenv, they all just called h5_cleanup
no matter. h5_cleanup now uses getenv_all to check the $HDF5_NOCLEANUP if
it is in parallel mode.
Platforms tested:
h5committested, tested pp in heping too.
Feature.
Description:
Modified it so that it can be compiled outside of HDF5 library as a standalone
program. e.g., mpicc -DSTANDALONE prog.c.
Platforms tested:
Tested in Red storm and heping.
Patch.
Description:
Copper would fail with a message of
0032-113 Out of memory in routine unknown, task 0
when run with 3 processes and size 1MB in MPI-IO tests.
It seems to be a copper MPIO error.
Solution:
Reduced the upper bound of default write size to 1/2MB (but
tests only go to 1/4MB) for now, pending permenant fix from
Copper.
Platforms tested:
Copper.
Minor bug fixes.
Description:
1. Changed free() calls to HDfree()
2. Corrected behavior of -m command-line parameter
3. Changed return value to always return 0.
Solution:
2. The -m flag tells the test to run only the MPI IO tests. However, it would incorrectly make
the test run both the MPI and POSIX tests (same as the default behavior).
3. This test is known to fail on many platforms, and, even on those platforms where it usually passes,
it is known to suffer transient failures (especially with small test file sizes). It's outcome is
also very dependent on the filesystem on which the testfile is created. Corrected
the program to always return success, so that it doesn't interfere with the daily tests.
The motivation for this is that
this test has nothing to do with the HDF library and is an auxiliary test. Failures in t_posix_complaint
do not necessarily mean that parallel HDF will fail, but simply indicate something to look into,
especially on new platforms. This is now an "output only" test, and any errors will be only be visible
in the output.
Platforms tested:
copper (all of these were minor changes)
Misc. update:
bug fix.
Description:
Fixed the segmentation fault errors in modi4, copper and tg-login.
It was due to the misuse of trying to realloc a pointer returned by
getenv_all. (not supposed to.)
Also rearranged the code so that option is checked first, then check
with environment variable, then use default setup. This saves the
need to do realloc at all.
Platforms tested:
Heping, modi4, shanti, copper (copper showed a different error now.)
Bug fix.
Description:
For some strange reason, getopt() does not appear to be defined in unistd.h
on colonelk when the source is compiled with -D_POSIX_SOURCE.
Solution:
Inserted some extern's for the missing variables to make the compiler happy.
Platforms tested:
colonelk, copper
Misc. update:
Feature
Description:
Added blurb about future todo's for this test.
Added support for HDF5_PARAPREFIX to determine the directory where the test file
is stored.
Solution:
Used getenv_all to get the value of HDF5_PARAPREFIX. Note that, if a command-line
parameter is passed to the program to specify a path, it will override the value of
HDF5_PARAPREFIX.
Platforms tested:
copper, colonelk
Misc. update:
Added t_posix_compliant to the rest of the build and patched up minor
compile bugs/warnings encountered on other platforms.
Description:
It seems that <getopt.h> needs to be included to get the file to build, even though the man page
seems to indicate that <unistd.h> should be sufficient.
Solution:
Platforms tested:
copper, colonelk, sol
Feature
Description:
Added posix compliance tests.
Solution:
These tests do increasingly complicated sets of writes followed by reads.
POSIX standards say that any read that can be proven to occur after a write
must include the data in that write. These tests attempt to verify whether the
underlying filesystem and i/o layer provide such guarantees.
Platforms tested:
copper, colonelk, red storm
Misc. update:
Bug fix
Description:
Fortran type generation was broken in two ways. Fixed both.
Solution:
Firstly, there were a couple of path problems. Fixed a typo and
specified the full path of a file.
Secondly, the dependencies weren't right when building with HDF5-specific
commands (make lib, make check-s, etc.). Tweaked dependencies
to fix the problem.
Platforms tested:
mir, modi4, sleipnir
Configure feature
Description:
Added 'make trace' target.
Solution:
Added tracing to 1.7. This was done automatically in 1.6, but left out
of 1.7 until now (oops!).
Tracing in 1.7 only happens manually, when the user types 'make trace.'
Tracing automatically requires more framework than it's worth.
I also fixed a couple of tracing bugs and ran trace.
Platforms tested:
mir, sleipnir, modi4
Misc. update:
Bug fix
Description:
Before this checkin, 'gmake check-s' would fail if there was a file in
the current directory named 'check-s'.
This is fixed under gmake (not sure how to fix for other makes).
Solution:
check, progs, install, etc. are what gmake calls "phony" targets,
which means that no file should be created. These targets can be
specified by a line of the form
.PHONY: check progs install ...
Automake adds this line for targets it knows about, but HDF5 has a
lot of custom rules. This checkin adds a .PHONY line for those rules.
I believe that only gmake recognizes the .PHONY line (at least, pmake
doesn't seem to), but a partial solution is better than none.
This error should occur very rarely anyway (the user has to manually
create files with names like 'build-check-s' or '_test').
Platforms tested:
mir, sleipnir, modi4
Bug fix/feature
Description:
Added support for -shlib in h5fc and h5c++.
Made check-install use -shlib when only shared libraries have been installed.
Solution:
h5fc and h5c++ didn't recognize -shlib. Stole code from h5cc to link against
shared libraries.
When static libraries are disabled, the examples Makefiles will automatically
use the -shlib option to link against shared libraries. Thus,
--disable-static and make check-install should work together.
Platforms tested:
heping(disable-static, enable-static, fortran, c++), modi4 (disable-static, fortran, c++, parallel, enable-static)
Bug fix
Description:
Failed parallel tests now cause make to exit with an error.
Solution:
Edited config/conclude.am to throw an error if parallel test programs fail.
Platforms tested:
heping, modi4
Adding comments and code clean-up for code that tests collective irregular selection
Description:
For better maintenance in the future, Add comments to list the number for
(start,count,block,stride) for irregular selection for effective testing
collective chunk IO feature development in the future.
Solution:
Platforms tested:
Linux 2.4(heping), mostly comments, no need to test on other platforms.
Misc. update:
Bug fix
Description:
Changed configure.in to use an environment variable TR to set the path
to the tr utility.
Solution:
There are two kind of tr on Solaris with slightly different syntax.
HDF5's configure relies on the "standard" tr. Traditionally, HDF5ers
have needed to make sure that the "right" tr was found before the
wrong one in their path; now they can use an environment variable.
Platforms tested:
mir, shanti, sol
Misc. update:
Forgot to update release notes. Off to do that now.
Code clean-up for collective regular chunk IO tests.
Description:
Add descriptions for each tests for future maintenance.
Solution:
Platforms tested:
Mostly comments, No need to use h5committest.
heping(linux 2.4)
Misc. update:
Makefile bug fix
Description:
Previously, automake didn't output rules to build perform/mpi-perf or
the test/gen_* programs.
Now these can be built by typing 'make mpi-perf' (or 'make foo') or by
configuring with --enable-build-all.
Solution:
Automake doesn't like having rules for programs it doesn't build. Tricked
it by having these programs built "sometimes"--whenever the user enables
--build-all. This should be used mostly for testing and to ensure that
these helper programs compile.
***IMPORTANT***
These programs do *not* currently compile. When --enable-build-all is used
(not the default), gen_new_fill fails because it uses an old API. This is
an existing "bug" that has simply been exposed by this checkin.
Platforms tested:
sleipnir, modi4, sol
Misc. update:
bug fix.
Description:
The complex derived datatype test assumed that the fill value would be 0. This is not
the case on all systems.
Solution:
Modified the test to check against a known value in the outbuf array, instead of the fill value.
Platforms tested:
heping and MCR.
Misc. update:
bug fix.
Description:
Fixed typo in a comment. The word "file" was supposed to be "fill." The explanation of
how the complex derived datatype test works is much clearer now.
Solution:
Platforms tested:
minor change.
Misc. update:
bug fix.
Description:
When a parallel test script test fails, make would continue because the
way it was setup inside a for loop. Fixed it by issuing an exit 1 inside
the loop.
There was also a typo error in the newer command comparision that it
must be $${chkname} in order to be valid. Also, the test script itself
was not checked in the newer lists. All fixed.
Platforms tested:
h5committested and also hand tested in heping pp mode.
Repair synchronization bug in the metadata cache in PHDF5
Also repair numerous other bugs that surfaced in testing the
bug fix.
Description:
While operations modifying metadata must be collective, we allow
independant reads. This allows metadata caches on different processes
to adjust to different sizes, and to place the entries on their dirty
lists in different orders. Since only process 0 actually writes
metadata to disk (all other processes thought they did, but the writes
were discarded on the theory that they had to be collective), this made
it possible for another process to modify metadata, flush it, and then
read it back in in its original form (pre-modification) form. The
possibilities for file corruption should be obvious.
Solution:
Make the policy that only process 0 can write to file explicit, and
visible to the metadata caches. Thus only process 0 may flush dirty
entries -- all other caches must retain dirty entries until they are
informed by process 0 that the entries are clean.
Synchronization is handled by counting the bytes of dirty cache entries
created, and then synching up between the caches whenever the sum
exceeds an (eventually user specified) limit. Dirty metadata creation
is consistent across all processes because all operations modifying
metadata must be collective.
This change uncovered may bugs which are repaired in this checkin.
It also required modification of H5HL and H5O to allocate file space
on insertion rather than on flush from cache.
Platforms tested:
H5committest, heping(parallel & serial)
Misc. update:
A bug fix
Description:
MPI_Status_IGNORE is treated as a NULL pointer for mpich 1.2.4 or similar MPI packages.
It caused segmentation fault for MPI derived datatype test.
Solution:
Define MPI_STATUS status,
and pass &status into MPI_File_read and MPI_File_write.
Platforms tested:
too trivial to test.
Misc. update:
Improvement.
Description:
The test may hang if there are system failures that some processors
are not working.
Solution:
Added the ALARM calls to limit all tests be done with the default
alarm time. So, even if a process is hanging, the ALARM signal
would terminate the process.
Platforms tested:
tg-login2 of NCSA.
Misc. update:
bug fix
Description:
the MPI_File_get_size returns a different value in one of the tests in Windows
comment the code and not run the test in windows
a ULL suffix on the harcoded return VRY return number is needed on AIX
Solution:
Platforms tested:
Windows
AIX
Misc. update:
new features
Description:
add support for compiling the library and testphdf5 in Windows
Solution:
Platforms tested:
Linux
AIX
Solaris
Windows VC6
Misc. update:
New feature.
Description:
Added the time command to the make check target to report time usage
of the execute of each test and test scripts. This gives us some idea
how long each test takes and some vague idea it is compute bound or
not.
powerpc-ibm-aix5.x:
Change $RUNPARALLEL default setting to allow it being invoked by the
time command.
Platforms tested:
h5committested.
Feature
Description:
Added H5_CFLAGS, etc. to 1.7 branch.
Now compilation flags can be put in H5_*FLAGS and they'll be used when
building hdf5 but not in h5cc.
Platforms tested:
mir, sleipnir, modi4
Misc. update:
Bug fix
Description:
Disabled C++ shared libraries for Sun Workshop compiler.
Solution:
This bug only seems to happen when using the -xarch=v9 flag to compile in
64-bit mode, but disabling shared libraries entirely for this compiler is
an easier fix (I don't know how to detect 64 bit mode from the command line).
The framework for disabling shared libraries for other C++ compilers is
in place.
Platforms tested:
sol, mir, sleipnir, modi4
Bug fix/feature
Description:
Disabled shared libraries for a number of Fortran compilers that don't
support them.
This allows other compilers to support shared Fortran libraries.
Solution:
Added a conditional, SHARED_FORTRAN_CONDITIONAL, which is true if
Fortran supports shared libraries. It is set in configure.in.
Platforms tested:
mir, sleipnir, colonelk, heping
Bug fix
Description:
Building with --disable-static seems to have been broken on several platforms.
Fixed this.
Configure will now disable shared libraries automatically when using pgf90
or when building on Cygwin.
Solution:
To avoid errors when using shared libraries with pgf90 (which I had not
realized were compiler-specific), I had manually set convenience libraries
to use the -static linking flag. Apparently, this is not necessary, and
caused these libraries to be created as empty archives when --disable-static
was used.
Also, some libraries were including other libraries, which was not
necessary.
I also fixed code in configure.in to correctly detect whether shared
or static libraries are enabled, and moved code that disables shared libraries
to before libtool is created (rather than editing libtool after the fact).
Despite repeated warnings that only one shared library can be linked at a
time, I have yet to have shared libraries break the linking of tests on
any system. We'll see if the Daily Tests turn up anything.
Platforms tested:
mir (fortran, C++), sleipnir (C++), modi4 (fortran, C++, parallel),
sol (fortran, C++), cygwin
Bug fix.
Description:
Some tests showed the filesize was not as expected. But the error was
intermittent. This was a racing condition as some processes finish
extend_chunked_dataset() sooner than others and return to the main
body which proceeds to call the next test which also uses the same
test data file and alters it. That messes up the "slower" processes
which then see unexpected filesize.
Also, the routine create_chunked_dataset() which creates test data file
actually was executed by all processes. That is wrong.
Solution:
Added a barrier at the end of extend_chunked_dataset to make sure all
processes are done with the test data file before returning.
Changed create_chunked_dataset such that only one process would create
the test data file. The rest does nothing but just wait for it to finish.
Platforms tested:
Tested in TG-NCSA in which the errors were detected.
Misc. update:
Bug fix#281
Description:
Committed a wrong copy in the previous checkin.
Solution:
Checked in the right one and did some code cleanup, rearrangment.
Platforms tested:
heping pp.
Misc. update:
tidy up.
Description:
MPE option created *.clog file whenever it is executed in MPI.
The cleanup of *.clog files were done in individual Makefile.in.
Often, it is forgotten.
Solution:
Moved the cleaning of *.clog files to CHECK_CLEANFILES in commence.am
so that it is applied whenever check-clean is called.
Platforms tested:
heping pp using MPE.
Bug fixes
Description:
This checkin fixes an occasional error on kelgia on sol during distclean.
It also causes test scripts to depend properly on the programs they're
supposed to be testing.
Solution:
The kelgia bug was due to some files being cleaned by automake and manually.
Removed the manual cleaning in src/Makefile.am.
Test script dependencies now need to be specified manually, since the
makefile can't guess what they test from their name. Currently all test
scripts in a given directory have a single list of dependencies--this was
easy and seems to be sufficient.
These dependencies are listed in the SCRIPT_DEPEND variable in the Makefile.am.
Platforms tested:
heping, mir, modi4, sol
Misc. update:
Bug #281
Description:
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. This was because parallel mode expects the allocation
mode be H5D_ALLOC_TIME_INCR only.
Solution:
Modified library to allocate more space when needed or directed if the
file is opened by parallel mode, independent of what the dataset allocation
mode is.
Platforms tested:
Heping pp.
minor tidy up.
Description:
Changed both test programs and test scripts to use the same suffixes
(.chkexe and .chklog). Changed from .log to .chklog to avoid running
into potential conflicts by other "things" (e.g., config.log).
Tested:
Heping.
bug fix
Description:
The previous version always tests with mpi-IO driver even when mpi-posix driver is set.
Solution:
Add the option to test mpi-posix driver for the first round collective chunk IO test.
Platforms tested:
IRIX64, AIX 5.1, Linux 2.4
Misc. update:
"bug fix"
Description:
The test_mpio_derived_dtype() often hangs when fails. So it was
not run by default to avoid hanging the daily tests or confusing
users. But then new system or new codes in collective chunk fails
while one can't tell for sure if it is because of the complicated
derived type failures or something else.
Solution:
Changed the logic so it is skipped only if it is known that
the complicated MPI derived type does not work. (This is
indicated by macro H5_MPI_COMPLEX_DERIVED_DATATYPE_WORKS NOT
defined.
Platforms tested:
heping pp (where it is tested by default).
Modi4 pp (where it is SKIPPED by default.)
I also forced modi4 to test it and modi4 said it actually
is working and should change the setting of
H5_MPI_COMPLEX_DERIVED_DATATYPE_WORKS to working?!!
Misc. update:
Code cleaning.
Description:
The block of code that is conditioned by the H5_MPI_COMPLEX_DERIVED_DATATYPE_WORKS
and manipulates the return code of test_mpio_derived_dtype, does not really
belong in main. If return code of test_mpio_derived_dtype needs to be
adjusted, it should be done in test_mpio_derived_dtype.
Solution:
Moved that block of code test_mpio_derived_dtype.
Platforms tested:
heping PP.
Misc. update:
Code cleanup
Description:
Trim trailing whitespace, which is making 'diff'ing the two branches
difficult.
Solution:
Ran this script in each directory:
foreach f (*.[ch] *.cpp)
sed 's/[[:blank:]]*$//' $f > sed.out && mv sed.out $f
end
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Modified collective chunk Io test
Description:
One test(collective chunk IO test 4) is no longer needed. Comment out first.
The code should be cleaned up later.
Solution:
Platforms tested:
Misc. update:
Bug fix and improvement.
Description:
The test output were not displayed for parallel tests.
The serial tests output were always displayed whether the test was
actually ran this time or not.
Solution:
Moved the display of test output inside the target where the actual
tests are run.
Platforms tested:
Tested in heping using pp and fortran.
