I'm having some odd problems with a single-threaded application that has a single file open and is writing to multiple data sets. Each data set has its own file hid_t, even though the file is the same. Are there any known problems with doing this?
What I'm seeing is:
1) hanging on reads. Stack:
[1] H5V_chunk_index
[2] H5D_create_chunk_file_map_hyper
[3] H5D_chunk_io_init
[4] H5D_read
[5] H5Dread
2) corruption of a data space (memory space, just before a write)
Chances are pretty good, I'd say, that I have another memory overwrite somewhere in my code that's causing this, but I thought I'd throw the question out there while I search.
Hi John,
I'm having some odd problems with a single-threaded application that has a single file open and is writing to multiple data sets. Each data set has its own file hid_t, even though the file is the same. Are there any known problems with doing this?
Nope, this should work.
What I'm seeing is:
1) hanging on reads. Stack:
[1] H5V_chunk_index
[2] H5D_create_chunk_file_map_hyper
[3] H5D_chunk_io_init
[4] H5D_read
[5] H5Dread
2) corruption of a data space (memory space, just before a write)
Chances are pretty good, I'd say, that I have another memory overwrite somewhere in my code that's causing this, but I thought I'd throw the question out there while I search.
Nothing comes to mind for that stack or behavior. Can you run with valgrind or some other memory debugger?
Quincey
···
On Aug 24, 2010, at 1:13 PM, John Knutson wrote:
Quincey Koziol wrote:
Hi John,
I'm having some odd problems with a single-threaded application that has a single file open and is writing to multiple data sets. Each data set has its own file hid_t, even though the file is the same. Are there any known problems with doing this?
Nope, this should work.
Nothing comes to mind for that stack or behavior. Can you run with valgrind or some other memory debugger?
Thanks for the response. The problem appears to have been user error (attempts to read/write from one data set using two completely different memory types). Copy/paste was the culprit, really.
Still, I did try and run the program using bcheck (maybe Sun-exclusive tool), and the HDF library would report an initialization failure and terminate when I did so, unfortunately.
Good to know my approach isn't fundamentally flawed, at least 
···
On Aug 24, 2010, at 1:13 PM, John Knutson wrote:
