Re: New PAGG patch for 2.6.10, new functionality

> I've noticed one minor issue with the implementation for skipping pagg
> associations during pagg_init.  If the register function finds that a
> task was taken off the task list during registration it traverses the
> list from the beginning.  Tasks that were skipped would therefore be
> looked at again.  Still, it's not a big issue.  I suppose clients
> should be able to handle looking at skipped tasks a few times.

Hi.  I was loooking at this a bit today.

I'm not quite sure how to improve this.  I could add a comment about this :)

I suppose we could have a list of already skipped tasks and not even
try them again if they were skipped once.  But I'm not sure if that is
too ugly?

Unless I hear more feedback, I'm just going to add a comment in the
existing comments for the init function pointer in pagg.h.  Something like:

The implementation of pagg_hook_register causes us to evaluate some tasks
more than once in some cases.  See the comments in pagg_hook_register for
why.  Therefore, if the init function pointer returns >0, which means that it
doesn't want a pagg association, that init function must be prepared to
possibly look at the same "skipped" task more than once.

--
Erik Jacobson - Linux System Software - Silicon Graphics - Eagan, Minnesota