Hey again - I was getting some strange segfaults when calling
H5Tconvert to go from unpacked to packed compound types (packed with
H5Tcopy + H5Tpack, while dealing with previously reported bug) in
another posix thread from where the H5 types were created/initalized.
The routines were crashing in H5TS_get_thread_local_value (rather
pthread_getspecific) - in H5TS_cancel_count_dec. Is this expected?
Is it not safe to use types allocated in the main thread in a slave
thread and does that make any sense if so? Is there any way to work
around the issue?
Can you give us an example program that we can use to reproduce this issue?
Please send it to the HDF Helpdesk.
Thanks!
-Barbara
ยทยทยท
========================================================
Barbara Jones, The HDF Group Helpdesk, help@hdfgroup.org
Support Services: https://www.hdfgroup.org/support/
-----Original Message-----
From: Hdf-forum [mailto:hdf-forum-bounces@lists.hdfgroup.org] On Behalf Of Jason Newton
Sent: Saturday, January 28, 2017 11:53 PM
To: hdf-forum
Subject: [Hdf-forum] H5Tconvert thread-related issue
Hey again - I was getting some strange segfaults when calling H5Tconvert to go from unpacked to packed compound types (packed with H5Tcopy + H5Tpack, while dealing with previously reported bug) in another posix thread from where the H5 types were created/initalized.
The routines were crashing in H5TS_get_thread_local_value (rather
pthread_getspecific) - in H5TS_cancel_count_dec. Is this expected?
Is it not safe to use types allocated in the main thread in a slave thread and does that make any sense if so? Is there any way to work around the issue?