Follow-up from "After a header merge the function pointers in the resulting matrix may still have inode versions"
The following discussion from !5505 (merged) should be addressed:
-
@knepley started a discussion: (+1 comment) @BarrySmith Is there any way that call goes inside the Merge? It seems like this could be a bug anytime we combine two AIJ matrices.
Yes put in a ops->postheadermerge function that is used for seqaij matrices.