[pgpool-committers: 2807] pgpool: Change implementation of serialize_accept to flock().

[Tatsuo Ishii]

Change implementation of serialize_accept to flock().

Former implementation uses semaphore. Problem is, when fail over
happens, pgpool child process exits without releasing semaphore
lock. There seems no reliable way to fix this. Locking with flock()
has benefit since the lock held by flock() is automatically released
before exiting the process. The lock file will be created as
"logdir/accept_lock".

Branch
------
master

Details
-------
http://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=cdf459f75d234426290e8ed0affc7b1bdc81839e

Modified Files
--------------
src/include/pool.h   |    3 +--
src/protocol/child.c |   58 ++++++++++++++++++++++++++++++++++++++++++++++++--
2 files changed, 57 insertions(+), 4 deletions(-)