As I already hinted a while ago, I am trying to resolve some of the
various shell scripts' obscure dependencies on various environment
variables and paths. My question is whether the use of the program
postconfig to determine the PGLIB value (or anything else, since there are
no checks done) is still encouraged/desired/done. To be clear: Contrary
to what some of you might start to believe, I do not want to remove every
little feature in PostgreSQL that I don't like/understand/use ;) I'm just
wondering.
My survey showed that the only two places where PGLIB is used is
createlang and initdb, so in a normal environment there is no good reason
to have PGLIB set all the time, anyway. In a developer's environment,
where these commands are executed many times, PGLIB is going to mess you
up if you want to install several versions, which is exactly the reason
why I am looking at this.
I found a pretty fool-proof (and with Tom's help portable) way to
determine the location of the needed files (the bki's and the PL handlers)
from the actual location of the shell script, and if that fails (which it
shouldn't) you can always use the --pglib/-L option. This will, unless you
intentionally use a particularly evil installation layout, ensure that any
initdb or createlang (or whatever else might use this) call is always
going to use the correct files and backend version without you having to
do any setting of anything (including PATH).
So, to summarize:
* PGLIB, keep it or lose it?
* postconfig, keep it or lose it?
--
Peter Eisentraut Sernanders väg 10:115
peter_e@gmx.net 75262 Uppsala
http://yi.org/peter-e/ Sweden
