pcp
[Top] [All Lists]

Re: python3 woes on f22

To: Nathan Scott <nathans@xxxxxxxxxx>
Subject: Re: python3 woes on f22
From: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Date: Tue, 14 Jul 2015 07:57:11 -0400
Cc: Mark Goodwin <mgoodwin@xxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <955515466.37555879.1436847748465.JavaMail.zimbra@xxxxxxxxxx>
References: <5594A1B5.5090003@xxxxxxxxxx> <y0mbnfhvx03.fsf@xxxxxxxx> <55A2FFBB.5000803@xxxxxxxxxx> <227885808.37023848.1436780775020.JavaMail.zimbra@xxxxxxxxxx> <y0my4ikulyq.fsf@xxxxxxxx> <955515466.37555879.1436847748465.JavaMail.zimbra@xxxxxxxxxx>
User-agent: Mutt/1.4.2.2i
Hi -

nathans wrote:
> [...]
> > But thinking of it as a user option (something changeable within
> > /etc/pcp.conf) is odd:
> 
> Not really, and pcp.conf is only one avenue of modifying it - its just
> like using /usr/bin/env to allow environment-driven alternatives. [...]

You elided the key part of my concern:

# this is a packaging decision (so dependencies can be declared &
# satisfied) rather than a sysadmin whim.

So, with choice of python interpreter being a run-time alternative,
how do you intend to address dependencies?  Should pcp.spec require
python or python3 & related libraries?  Sure, you can make it
by-default-match the one that pcp.conf will be preconfigured for.  But
if the sysadmin changes it, the software will break (since the other
python{3,2} dependencies are no longer assuredly satisfied by the
packaging system).

I suggest against making this setting run-time configurable a la
pcp.conf or the environment, and instead hard-coding it within the
wrapper.

- FChE

<Prev in Thread] Current Thread [Next in Thread>