-
Notifications
You must be signed in to change notification settings - Fork 68
io/ompio: Add barrier to file_close and file_set_size #1302
Conversation
This fixes a bug reported on the mailing for ompio. https://www.open-mpi.org/community/lists/users/2016/05/29333.php
|
Test FAILed. |
|
the same hcoll issues on mellanox 19:54:07 --- MCA component coll:hcoll (m4 configuration macro) |
|
Test FAILed. |
|
@edgargabriel please assign a reviewer. |
|
I actually did, butit does not show up for whatever reason, lets try again |
|
bot:assign:@ggouaillardet |
|
Test FAILed. |
|
bot:mellanox:retest |
1 similar comment
|
bot:mellanox:retest |
|
Test PASSed. |
|
👍 |
|
@jsquyres ready to go |
This fixes a bug reported on the mailing for ompio.
https://www.open-mpi.org/community/lists/users/2016/05/29333.php
:bot:milestone:v2.0.1
:bot🏷️bug
:bot:assign:@ggouaillardet
@ggouaillardet would you mind reviewing this fix, since you did the counterpart in romio314 ? The issue here was that the barrier in file_close was not sufficient, there has to be a barrier in file_set_size as well, since the user test case otherwise had the potential to reset the size to 0 after some processes already wrote some data. It works now for me for both file_write_at and file_write_shared without any issues.