opal: Segfault avoidence in class and mca/btl/ofi #10466
Conversation
Fixed a loop causing segfaults in opal_object.h and fixed goto fail block to check for NULL before potential dereference as we go there for NULL in what we dereference Signed-off-by: Josh Fisher <[email protected]>
|
Can one of the admins verify this patch? |
|
refers to #10370 |
|
ok to test |
|
The IBM CI (GNU/Scale) build failed! Please review the log, linked below. Gist: https://gist.github.com/26548c94bf5feedcb59c28466652fc15 |
|
The IBM CI (XL) build failed! Please review the log, linked below. Gist: https://gist.github.com/e30e7d3f3c8a8ab3c7cd48513150d654 |
|
Looks like I have some more work to do here, let me keep working on this. |
|
The IBM CI (PGI) build failed! Please review the log, linked below. Gist: https://gist.github.com/3e5d7ae58474e5874cbeea704cac8078 |
| @@ -468,7 +468,7 @@ static inline void opal_obj_run_destructors(opal_object_t *object) | |||
| assert(NULL != object->obj_class); | |||
|
|
|||
| cls_destruct = object->obj_class->cls_destruct_array; | |||
| while (NULL != *cls_destruct) { | |||
| while (NULL != cls_destruct) { | |||
There was a problem hiding this comment.
How it is possible to have a zero-level class hierarchy for any object other than opal_class_t ? It looks as if you are trying to free a class that was not yet initialized.
|
Github pro tip: You don't need to close a PR to modify it. |
|
Yeah, I wasn't sure how you guys do it and was hoping to not waste anyone's time on things that definitely won't work. |
|
Create the PR as draft, and keep it that way until completion. You can still request reviews and have discussions in the PR, but it will not be merged until you get it out of draft mode. |
Fixed a loop causing segfaults in opal_object.h and fixed goto fail
block to check for NULL before potential dereference as we go there
for NULL in what we dereference
Signed-off-by: Josh Fisher [email protected]