Running hdf5-1.8.6 with the downloaded HDFView 2.7 binary for linux-x86_64 on CentOS 5.4 as well as the HDFView 2.7 src compiled locally. Am seeing a crash while opening a packet table. The h5 is attached as PacketTableIssue.h5 and the java crash report is also attached.
The table is /rec/Events and its properties can be viewed in HDFView 2.7. It can be read by C++ code and h5dump reads and shows the table.
Roger, thank you for reporting the problem. we will take a look at it. --pc
···
On 4/6/2011 3:04 PM, Roger Martin wrote:
Hi,
Running hdf5-1.8.6 with the downloaded HDFView 2.7 binary for linux-x86_64 on CentOS 5.4 as well as the HDFView 2.7 src compiled locally. Am seeing a crash while opening a packet table. The h5 is attached as PacketTableIssue.h5 and the java crash report is also attached.
The table is /rec/Events and its properties can be viewed in HDFView 2.7. It can be read by C++ code and h5dump reads and shows the table.
We have fixed the crash at our current hdf-java source repository.
I think you used H5Tvlen_create(H5T_C_S1) to create variable length
strings in the packet table, which may cause problems. A better way to
create variable length strings is to use H5T_C_S1 directly, e.g.
int tid = H5Tcopy(H5T_C_S1);
H5Tset_size(tid, H5T_VARIABLE);
(HDFView works fine if strings are created in this way)
Thanks
--pc
···
On 4/6/2011 3:04 PM, Roger Martin wrote:
Hi,
Running hdf5-1.8.6 with the downloaded HDFView 2.7 binary for linux-x86_64 on CentOS 5.4 as well as the HDFView 2.7 src compiled locally. Am seeing a crash while opening a packet table. The h5 is attached as PacketTableIssue.h5 and the java crash report is also attached.
The table is /rec/Events and its properties can be viewed in HDFView 2.7. It can be read by C++ code and h5dump reads and shows the table.
And you're right; I did use the H5Tvlen_create(H5T_C_S1). I'll switch to your better way. This may also help out python scripting using h5py/numpy on these same packet tables. I'm having good success with hdf5 not only with data but as the persistence bridge between languages and software integrations.
To the whole HDF team, thank you
···
On 04/07/2011 04:53 PM, Peter Cao wrote:
Hi Roger,
We have fixed the crash at our current hdf-java source repository.
I think you used H5Tvlen_create(H5T_C_S1) to create variable length
strings in the packet table, which may cause problems. A better way to
create variable length strings is to use H5T_C_S1 directly, e.g.
int tid = H5Tcopy(H5T_C_S1);
H5Tset_size(tid, H5T_VARIABLE);
(HDFView works fine if strings are created in this way)
Thanks
--pc
On 4/6/2011 3:04 PM, Roger Martin wrote:
Hi,
Running hdf5-1.8.6 with the downloaded HDFView 2.7 binary for linux-x86_64 on CentOS 5.4 as well as the HDFView 2.7 src compiled locally. Am seeing a crash while opening a packet table. The h5 is attached as PacketTableIssue.h5 and the java crash report is also attached.
The table is /rec/Events and its properties can be viewed in HDFView 2.7. It can be read by C++ code and h5dump reads and shows the table.
FYI: we are going to add more information to the reference manual for H5Tvlen_create()
so that users will not be confused.
Thanks
--pc
···
On 4/7/2011 4:32 PM, Roger Martin wrote:
Thank you, Peter!
And you're right; I did use the H5Tvlen_create(H5T_C_S1). I'll switch to your better way. This may also help out python scripting using h5py/numpy on these same packet tables. I'm having good success with hdf5 not only with data but as the persistence bridge between languages and software integrations.
To the whole HDF team, thank you
On 04/07/2011 04:53 PM, Peter Cao wrote:
Hi Roger,
We have fixed the crash at our current hdf-java source repository.
I think you used H5Tvlen_create(H5T_C_S1) to create variable length
strings in the packet table, which may cause problems. A better way to
create variable length strings is to use H5T_C_S1 directly, e.g.
int tid = H5Tcopy(H5T_C_S1);
H5Tset_size(tid, H5T_VARIABLE);
(HDFView works fine if strings are created in this way)
Thanks
--pc
On 4/6/2011 3:04 PM, Roger Martin wrote:
Hi,
Running hdf5-1.8.6 with the downloaded HDFView 2.7 binary for linux-x86_64 on CentOS 5.4 as well as the HDFView 2.7 src compiled locally. Am seeing a crash while opening a packet table. The h5 is attached as PacketTableIssue.h5 and the java crash report is also attached.
The table is /rec/Events and its properties can be viewed in HDFView 2.7. It can be read by C++ code and h5dump reads and shows the table.
