Re: CVS commit: pkgsrc/devel/glib2

[Roy Marples <roy%NetBSD.org@localhost>]

On 19/04/2011 17:34, Alistair Crooks wrote:
> Let's discuss the embedded system, then, since there's a fairly high
> probability that glib2 will be installed on it.  Is there a good
> reason there's now a python interpreter shipped with it?  I'm not sure
> "developer laziness" or "more correct" are good enough reasons for me,
> which make up for the increased space required, or functionality to
> secure.

Embedded systems are irrelevant to the topic at hand.
They might be pertinent to you and Gregg.

What is relevant is that pkgsrc should not install broken things, which 
what I'm addressing here.

> There must be a reason that dhcpcd-ui is shipped as a different part
> of the system to dhcpcd, to take one very useful utility at random
> (well, almost).

You, as well as I, know the reason is dhcpcd is like that is to make 
packaging and dependencies easier. That is an *upstream* decision I made.

> But, yes, if we absolutely, unarguably, non-negotiably need a python
> interpreter as an intrinsic, atomic, inherent part of glib2, let's
> ship it.  What is the glib2 piece (i.e.  the user library piece) that
> needs this in order to function?

The *upstream* glib2 decision is to install, without option, a part that 
absolutely, unarguably, non-negotiably need a python interpreter.
As to the need of gtester-report (the glib2 piece that depends on 
python) I don't know. I personally don't use it, nor care for it.

What I do care about is adding qa checks to ensure broken packages are 
not installed. If you embedded guys care about reducing the need of 
dependencies then more power to you as less bloat is always good.

Thanks

Roy