If you have some time to test this within the next two weeks, we would
greatly appreciate it, as while we test on the platforms we have access
to, there are many more that we are unable to test on ourselves. Please
report any errors found by October 10th. If no critical errors are
reported beforehand, we're targeting mid-October for our release date.
If you have some time to test this within the next two
weeks, we would
greatly appreciate it, as while we test on the platforms we
have access
to, there are many more that we are unable to test on
ourselves. Please
report any errors found by October 10th. If no critical
errors are
reported beforehand, we're targeting mid-October for our
release date.
Thank you!
The HDF Team
on cygwin-1-7.8s (snapshots 20100924) with gcc-4.3.4
with shared libs, hdf5 builds fine and pass almost
all the tests.
Only test failures :
···
============================
cache_api.exe Test Log
Testing MDC/FAPL related API calls PASSED
Testing MDC/FILE related API calls *FAILED*
check_file_mdc_api_calls(): failure_mssg = "Unexpected internal config #3.".
Testing MDC API smoke check -SKIP-
Long tests disabled.
Testing MDC/FAPL related API input errors PASSED
Testing MDC/FILE related API input errors HDF5-DIAG: Error detected
in HDF5 (1.8.6-pre1) thread 101384248: #000: ../../hdf5-1.8.6-pre1/src/H5F.c line 1427 in H5Fcreate(): unable to create file
major: File accessability
minor: Unable to open file #001: ../../hdf5-1.8.6-pre1/src/H5F.c line 1230 in H5F_open(): unable to truncate a file which is already open
major: File accessability
minor: Unable to open file
*FAILED*
check_file_mdc_api_errs(): failure_mssg = "H5Fcreate() failed.
...
Testing query functions of compiler conversion *FAILED*
Can't query conversion function
...
Check file ../src/.libs/libhdf5.so strings: '../src/.libs/lib
hdf5.so': No such file
FAILED
I tried a plain old ./configure & make on Mac OS X 10.6.4 with Xcode
3.2.4 (using default gcc 4.2) and it built, but there were an ungodly
number of warnings, mostly "warning: will never be executed". As I
said, I didn't set any particular flags. 'make check' reported no failures.
If you have some time to test this within the next two weeks, we would
greatly appreciate it, as while we test on the platforms we have access
to, there are many more that we are unable to test on ourselves. Please
report any errors found by October 10th. If no critical errors are
reported beforehand, we're targeting mid-October for our release date.
--
____________________________________________________________
Sean McBride, B. Eng sean@rogue-research.com
Rogue Research www.rogue-research.com
Mac Software Developer Montréal, Québec, Canada
Thanks to everyone who tested 1.8.6-pre1. We greatly appreciate your effort.
The release date was pushed to November due to the issues discovered during the testing. We have been working on the fixes and will make the next pre-release candidate available for testing as soon as possible. Stay tune.
If you have some time to test this within the next two weeks, we would
greatly appreciate it, as while we test on the platforms we have access
to, there are many more that we are unable to test on ourselves. Please
report any errors found by October 10th. If no critical errors are
reported beforehand, we're targeting mid-October for our release date.
If you have some time to test this within the next two
weeks, we would
greatly appreciate it, as while we test on the platforms we
have access
to, there are many more that we are unable to test on
ourselves. Please
report any errors found by October 10th. If no critical
errors are
reported beforehand, we're targeting mid-October for our
release date.
Thank you!
The HDF Team
on cygwin-1-7.8s (snapshots 20100924) with gcc-4.3.4
with shared libs, hdf5 builds fine and pass almost
all the tests.
Only test failures :
============================
cache_api.exe Test Log
Testing MDC/FAPL related API calls PASSED
Testing MDC/FILE related API calls *FAILED*
check_file_mdc_api_calls(): failure_mssg = "Unexpected internal config #3.".
Testing MDC API smoke check -SKIP-
Long tests disabled.
Testing MDC/FAPL related API input errors PASSED
Testing MDC/FILE related API input errors HDF5-DIAG: Error detected
in HDF5 (1.8.6-pre1) thread 101384248: #000: ../../hdf5-1.8.6-pre1/src/H5F.c line 1427 in H5Fcreate(): unable to create file
major: File accessability
minor: Unable to open file #001: ../../hdf5-1.8.6-pre1/src/H5F.c line 1230 in H5F_open(): unable to truncate a file which is already open
major: File accessability
minor: Unable to open file
*FAILED*
check_file_mdc_api_errs(): failure_mssg = "H5Fcreate() failed.
...
Testing query functions of compiler conversion *FAILED*
Can't query conversion function
...
Check file ../src/.libs/libhdf5.so strings: '../src/.libs/lib
hdf5.so': No such file
FAILED
If you have some time to test this within the next two weeks, we would
greatly appreciate it, as while we test on the platforms we have access
to, there are many more that we are unable to test on ourselves. Please
report any errors found by October 10th. If no critical errors are
reported beforehand, we're targeting mid-October for our release date.
Hi,
I tried a plain old ./configure & make on Mac OS X 10.6.4 with Xcode
3.2.4 (using default gcc 4.2) and it built, but there were an ungodly
number of warnings, mostly "warning: will never be executed". As I
said, I didn't set any particular flags. 'make check' reported no failures.
yah, we need to fix it. For this release we worked hard to remove all memory leaks. Hope to tackle warnings for the next one.
I am writing to update you on the HDF5 1.8.6 release.
The release is on hold due to tests failures in the parallel HDF5 library. We still do not know if there is a problem in HDF5, in the version of MPI I/O library we are using, or in both.
We now have a standalone HDF5 program (attached) that consistently reproduces the problem on NCSA's abe with Lustre and MVAPICH2. The data in the created file is not correct. There is no problem when OpenMPI is used.
While we are working on the MPI counterpart of the program, we will highly appreciate if you can test the program with the parallel HDF5 libraries installed on your systems and report
- testing results
- version of HDF5 (any of the 1.8.* series will be OK to use)
- version of the MPI I/O library
- version of OS and file system
- version of compilers used to build MPI I/O and HDF5
You should be able to compile the program by using h5pcc compiler script (usually it can be found in the bin directory under the HDF5 installation directory). The program should be run on several nodes with the number of processes greater than number of nodes, for example, 4 nodes and 6 processes.
I tried a plain old ./configure & make on Mac OS X 10.6.4 with Xcode
3.2.4 (using default gcc 4.2) and it built, but there were an ungodly
number of warnings, mostly "warning: will never be executed". As I
said, I didn't set any particular flags. 'make check' reported no
failures.
yah, we need to fix it. For this release we worked hard to remove all
memory leaks. Hope to tackle warnings for the next one.
Why do you enable all those warning flags then?
···
On Mon, 27 Sep 2010 19:40:43 -0500, Elena Pourmal said:
--
____________________________________________________________
Sean McBride, B. Eng sean@rogue-research.com
Rogue Research www.rogue-research.com
Mac Software Developer Montréal, Québec, Canada
On Mon, 27 Sep 2010 19:40:43 -0500, Elena Pourmal said:
I tried a plain old ./configure & make on Mac OS X 10.6.4 with Xcode
3.2.4 (using default gcc 4.2) and it built, but there were an ungodly
number of warnings, mostly "warning: will never be executed". As I
said, I didn't set any particular flags. 'make check' reported no
failures.
yah, we need to fix it. For this release we worked hard to remove all
memory leaks. Hope to tackle warnings for the next one.
Why do you enable all those warning flags then?
We use them during development, so that we are aware of the possible issues. However, as you mention, they are distracting and should be optional. As Elena mentions, we're planning to modify our configure scripts to add this option.