Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Most likely, we would define a port name in configure.in and test for
> that before accepting that flag. Also, there would be _no_
> documentation of this flags existance in the PostgreSQL documentation.
> It is only used by the postmaster calling postgres, so that should be
> OK.
I concur with Peter: there seems to be no need for this hack at all.
Just install two copies of the postmaster/postgres executable --- ie,
use cp instead of ln on platforms where ln doesn't work. The postgres
executable would only be used for standalone-backend operation, which'd
probably mean you could remove it after initdb, if the disk space hit
bothers you.
regards, tom lane