On 01/21/10 05:13 PM, [email protected] wrote:
Folks,
I have a small webrev that fixes a pair of locking bugs, as well as a
performance problem that Danek found on Friday.

These are the bugs:

        3720 Race condition exists for cancelation using the API
        13925 installs slow down towards the end
        14049 Classes that handle RLock have trouble with NRLock

This is the webrev:

        http://cr.opensolaris.org/~johansen/webrev-3720/

Seems reasonable, although I wonder if the changes for 13925 made any measurable impact on our current RSS usage?

As for the cancel changes, the only comment I have there is that it seems like (excluding the handler/cleanup functions in the API) every caller that wants to enable cancel has to through the acquire/release lock dance.

Perhaps we should just have a __enable_cancel function as we already have a __disable_cancel that performs the same task? Or consolidate them somehow into __set_can_be_canceled or something else?

Cheers,
--
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to