[cle-release-team] kernel issues

[David Horwitz]

To move my comments from the Jira.

I am maintaining a -1 on any changes to sqlService transaction handling 
unless:
- the change includes unit tests and/or
- its demonstrated that the transaction handling is covered by existing 
unit tests

KNL-799 includes a -1 from Stephen
KNL-815 inlcudes a -1 from Steve Swinsber and myself



D

On 01/20/2012 07:14 PM, Charles Hedrick wrote:
> Kernel issues from yesterday:
>
> 799: I've added a comment. Probably not critical to do, but I recommend setting rollbackOnReturn in trunk so it is tested for the next release.
>
> 797: was resolved by reverting 433. Should be fine as long as no one puts 433 back.
>
> 815: we have been running the latest patch in production. I can't prove that the race condition still exists since we upgraded mysql without more work than I current have time to do. It doesn't look like there is an overt deadlock in the code. I think the problem is that subselects cause locks to be generated implicitly, and there's no clear documentation on what locks happen. With Mysql at least, they are version-specific. I'm going to leave the patch in the Rutgers kernel, as I believe it's safer to do the lock. Our problem seemed to have been provoked by Lesson Builder. I modified the LB code to make sure that a lock happens even in an unmodified kernel.
>
> _______________________________________________
> cle-release-team mailing list
> cle-release-team at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/cle-release-team
>