In current sources, lock.c's LockDisable() function is called only
for bootstrapping or if the (useless) -L backend switch is used.
I have verified that it's not needed for bootstrapping: initdb
succeeds just fine without it. Accordingly, I'm strongly tempted
to remove the function, the switch, and the tests for LockingDisabled()
in all the lock-related functions. Does anyone see a reason to
keep 'em?
regards, tom lane