I hadn’t noticed until recently that --enable-rpath is passed to configure on Red Hat PostgreSQL builds and inherited by all extensions (via pg_config) on that platform. rpmlint complains about it on several Red Hat variants (well, in fact it _errors_).
Poking around mailing lists and git history, I was able to ascertain that it wasn’t always this way, with some caveats around how plperl and plpython are build on these platforms.
Does anyone have a pointer as to why this was changed for PostgreSQL itself? Or when?
Do extensions need to worry about this? I see some that use e.g. chrpath on binaries, but without understanding the reason PostgreSQL itself uses an rpath-enabled build I didn’t want to add that.
Planning on just filtering this rpmlint error for now, but I am curious for the motivation.
--
Jason Petersen
Software Engineer | Citus Data
303.736.9255