Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> Peter Eisentraut wrote:
>> I figured it would make sense if pgarch.c used the same mechanism that
>> postmaster.c uses to report the various variants of regular and signal
>> exits.
> Hmm. Getting rid of the "(PID 0)" is going to be a mess enough for
> translations that I think it is worth pgarch.c having its own routine
> for this. Furthermore I think the detailed archive command should be
> reported in an errdetail() field, which makes it even farther off.
I agree with Alvaro --- trying to make LogChildExit serve two masters
will be uglier and less maintainable than having two copies of the
not-really-so-complex logic involved. Leave postmaster.c alone and
just make pgarch.c smarter.
(But having said that, +1 for improving the message.)
regards, tom lane