On Mon, Nov 21, 2016 at 11:48 PM, Thomas Munro
<thomas.munro@enterprisedb.com> wrote:
> Here's a version that works that way, though it allows you to call
> ConditionVariablePrepareToSleep *optionally* before you enter your
> loop, in case you expect to have to wait and would rather avoid the
> extra loop. Maybe there isn't much point in exposing that though,
> since your condition test should be fast and waiting is the slow path,
> but we don't really really know what your condition test is. I
> thought about that because my use case (barrier.c) does in fact expect
> to hit the wait case more often than not. If that seems pointless
> then perhaps ConditionVariablePrepareToSleep should become static and
> implicit. This version does attempt to suppress spurious returns, a
> bit, using proclist_contains. No more cvSleeping.
This version looks good to me and I have committed it after doing a
bit more work on the comments.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company