[Date Prev][Date Next]
(ITS#4343) WAKE_LISTENER loses wake events
Full_Name: Howard Chu
Submission from: (NULL) (126.96.36.199)
Submitted by: hyc
This is already fixed in HEAD, just noting it here.
I recall we discussed the possibility of this being a problem in the past.
Apparently the last round if ITS#3950 changes caused it to occur more
frequently. The result being that slapd sits idle while clients wait for
WAKE_LISTENER needed a counter in non-threaded slapd to prevent the wake pipe
from filling up. But with threaded slapd, there's a race between zeroing the
counter and reading the wake pipe, so it's best to just write the pipe
unconditionally and dispense with the counter.