pg_basebackup contains an option called --write-recovery-conf allowing the user to generate automatically some values for recovery.conf. As in this patch all the recovery parameters are moved to postgresql.conf as GUCs, I replaced the existing option by a new option called --write-enable-standby that simply creates an empty standby.enabled file in the base backup if specified. By looking at the code of pg_basebackup, I noticed that adding some automatically-generated input to postgresql.conf would not impact so much the code when using plain format. However adding special handling in the code for tar format would make the code less readable and less intuitive. Also do we really need to generate recovery parameters with pg_basebackup? In the spec discussed we want the user to define once recovery parameters in the master's postgresql.conf such as the same values could be reused directly in the slave's base backup.
I have also reworked the documentation a bit: - Addition of a section called "Recovery" in the configuration section to describe all the parameters that are activated only if standby.enabled is found at server startup. - Addition of a note at the end of the new "Recovery" section indicating how to migrate an existing configuration using recovery.conf to the new system using include_dir or include_if_exists.
