Ferdinand Smit <ferdinand@telegraafnet.nl> writes:
>> I do see an issue here, which is that the STATISTICS messages should be
>> emitted by elog() so that they will be tagged with a PID when log_pid is
>> true. But that's easily fixed.
> Second try, this time by using the elog function. Is this a better solution ??
I've already committed changes that do the former (emit the STATISTICS
messages via elog, not fprintf, so that they can be timestamped and
PID-stamped). Does that address your concern?
regards, tom lane